vol 2 no 4

(IJCNS) International Journal of Computer and Network Security, Vol. 2, No. 4, April 2010

1

RD-Optimisation analysis for H.264/AVC scalable video coding

Sinzobakwira Issa1, Abdi Risaq M. Jama2 and Othman Omar Khalifa3

1 Olympia College, School of Engineering Persiaran Raja Chulan, 50200 Kuala Lumpur, Malaysia

[email protected]

2, 3 International Islamic University Malaysia, Department of Electrical and Computer Engineering Jalan Gombak, Box: 53100 Kuala Lumpur, Malaysia

Abstract: The development of multimedia propagations and applications has led to a greater expansion in the field of video transmission over a heterogeneous media as well as iterative delivery platforms with dedicated content requirements. It is known that conventional video coding systems encode video content with given bitrates adapted to a specific function or application. As a result, conventional video coding does not meet the fundamental requirements of the state-of-the-art flexible digital media application. The newly technology based on scalable video coding appears as a new modus operandi that has the ability to satisfy the underlying requirements. In this work, a multi-users scenario was considered for an optimum performance between multiple streams. A rate distortion optimized video frame dropping strategy which can be applied on active network nodes during high traffic intensity was developed. The concept of scalability here, come to introduce the operability of high level of suppleness coding and decoding systems. A base layer which can display the suitable quality of the premium file was considered and take care of the improvement of video quality.

Keywords: Bitrates, PSNR, bandwidth, multi-users scenario and RDO.

1. Introduction The past few decades, starting in the early nineties, a remarkable development has been achieved in the field of video compression. A lot of efforts were and still are being exerted for compressing, storing data in digital medium and allocation over the web.

It is very crucial to have the idea of monochrome digital video data sequence which is a set of individual pictures called frames happening at predetermined time increments. This frame needs to be considered as a light intensity of two dimensions in terms of function of variable x and y; f(x, y), where x and y denote special coordinates and the value off at any point (x, y) is proportional of the brightness of the frame or the gray level at the point for monochrome. The normal standard speed at which these frames are displayed is 30 frames per second.

This representation is called canonical representative. However, this canonical representation has negative impact because it needs very huge amounts of memory, resulting in impracticality of being stored or shared on the web or to be launched into digital channel. The fact may seem as an amusing game when we try to illustrate how it could be done.

The clear picture is an example of a 100 minutes movie displayed at 30 frames per second with width of frame 640x480 pixels with each pixel taking 3 bytes of memory.

The reality shows that, for each second of the movie, the requirement be at least 27MB of memory; as a result, the entire movie will need almost 162GB of memory. If this movie were stored on DVD’s, then considered the current DVD capacity of 4.7 GB, would roughly require 35 DVD’s. Therefore, video needs to be compressed considerably for efficient storage and sharing over the web [1] However, there are a lot of redundancies within the video data that can be eliminated yielding file size reduction or compression.

2. H.264/AVC Scalable Video Coding

2.1 Basic H.264/AVC structure The H.264/AVC standard has a range of coding tools contributing to its high compression performance, flexibility and robustness. However, the performance improvements come at a cost of significantly high computational complexity. Therefore, encoder implementations should make use of the available coding tools effectively to achieve the desired compression performance with the available processing resources.

H.264/AVC is an extremely scalable video codec, delivering excellent quality across the entire bandwidth spectrum, from high definition television to the video conferencing and 3G mobile multimedia. The following can thusly be summarized as the important differences.

• Enhanced motion prediction capability • Use of a small block-size exact match transform • Adaptive in-loop deblocking filter • Enhanced entropy coding methods

Figure 1. H.264/AVC structure

2.2 Scalable Video Coding


2

Scalable video coding is desirable in heterogeneous and error-prone environments for various reasons. For example, scalable coding helps streaming servers avoid congestions in network by allowing the server to reduce the bitrate of bitstreams whilst still transmitting a useable bitstream.

One application for scalability is to improve error resilience in transport systems that allow different qualities of service.

For example, the essential information could be delivered through a channel with high error protection. Scalability can also be used to enable different quality representations depending on playback devices processing power.

Devices with better processing power can decode and display the full quality version, whereas devices having lower processing power decode the lower quality version.

2.3 Types of SVC There are three conventional types of scalability: temporal, quality and spatial. Temporal scalability enables adjustment of picture rate.

a) This is commonly carried out with either disposable pictures or disposable sub-sequences, which are explained later on. Picture rate adjustment is then simply done by removing these disposable parts from the coded sequence thus lowering the frame rate.

b) In conventional quality scalability, also known as SNR scalability, an enhancement layer is achieved with pictures having finer quantizers than the particular picture in the lower reference layer[3]. In coarse-granularity quality scalability, pictures in enhancement layers may be used as prediction references and therefore all the enhancement layer pictures in a group of pictures typically have to be disposed as a unit. In fine granularity scalability, the use of enhancement layer pictures as prediction sources is limited and therefore finer steps of bitrate can be achieved compared to coarse-granularity scalability.

c) Finally, spatial scalability is used for creation of multi-resolution bitstreams to meet different display requirements or constraints and is very similar to SNR scalability [5].

A spatial enhancement layer enables recovery of coding loss between an up-sampled version of the reconstructed layer used as a reference by the enhancement layer and a higher resolution version of the original picture. 3. Rate Distortion Optimization

3.1 Lagrangian multiplier method In H.264/AVC, it is the art of the encoder to have the ability of having the effective way of encoding a given video sequence by selecting among numerous ranges of modes and parameters.

The encoder targets to achieve optimum rate distortion performance by choosing the best of modes and parameters of a given video. Doing this, the encoder would be looking to minimize distortion in a sequence of particular video.

Rate-distortion Optimisation (RDO) methods used in video compression are discussed in [6] [2], which include dynamic programming and Lagrange optimisation methods.

A Lagrange optimisation method, which is also known as Lagrange multiplier method, offer computationally less complex (although sometimes sub-optimal) solutions to the optimisation problem was proposed. Due to its less complex nature, a specific form of the Lagrange optimisation method has been used in rate-distortion optimisation of H.264/AVC [10]. 3.2 Constrained Optimisation Problem The objective function within source constraints is minimized or maximized by the constrained optimization. In the case here of video coding standards, this issue of constrained optimization can be considered as reducing the amount of distortion of a given video sequence meaning to strive looking to increase the number of bits that can be encoded in exactly that particular coding sequence[4]. Below is the mathematical representation of the constrained optimization unit; Let S represent all the allowable vectors and let B an element of S, (BЄS). The objective function is defined for all B in S as D(B) and the constraint function R(B) is defined for all B in S . The constrained problem can be presented as: Given a constraint Rc, find

BЄS Subject to

The solution (BЄS*) to the problem satisfies that R (B*) ≤ Rc and D (B*) ≤D (B) for all B In S*, where

That is, if the solution to the problem is B*, then there is no other B in S which satisfies the constraint Rc, that will result in a smaller value for the objective function than D (B*). The Lagrange multiplier theory offers a way of solving the above constrained problem (i.e. finding B*) by representing the problem as an unconstrained problem [3]. 3.3 Major Theorem The constrained optimisation problem was presented earlier in previous section, equation (2). The Lagrange theory represents the constrained problem as an unconstrained problem as follows: Theorem: for any λ≥0, the solution B*(λ) to the unconstrained problem

This is considered as solution of the constrained problem in (1) presenting Rc = R (B* (λ)) as the constraint. Proof of the theorem


3

If B* (λ) is the solution to the constrained problem (4) then:

Therefore,

If this is true for all B in S, it is true for a subset of B in S where,

Now, for the above subset and for any λ≥0:

Therefore with the constraint Rc =R (B* (λ)), the solution B* for the unconstrained problem is also the solution for the constrained problem.

It should be noted that the theory does not guarantee a solution for the constrained problem. It only states that for any λ≥0 of the unconstrained problem, there is a corresponding constrained problem which has the same solution as the unconstrained problem. 3.4 Optimisation problem Consider a macroblock, for which the encoder can encode the macroblock using only one of the ‘K’ possible modes given by the set m = {m1,, m2, … , mK }. Let ‘M’ (M Єm) be the mode selected to code the macroblock. In the context of H.264/AVC, these mode allocations could be any allowable combination of macroblock partition modes, Quantisation Parameters (QP), choice of reference frames etc… so that the K possible modes will include all the possible admissible parameter combinations for the Macroblock

Define the objective function D(M) and constraint function R(M) , where D(M) and R(M) are distortion and rate of the macroblock as a result of selecting a particular coding mode. If the rate constraint is Rc, the constraint problem is defined as: Find the coding mode M*,

Subject to

This may be written as an unconstrained problem using a Lagrange multiplier:

Where the solution to (4-11), M*, would satisfy,

The optimum coding mode M* (if one exists) can be found by solving (14). That means, when the macroblock is coded in mode M* it would satisfy the target rate (R (M*) = Rc). All the other modes (if they exist) that satisfy R (M)≤ Rc will have a higher distortion than D(M*).

The term D (M) + λ .R (M) in equation (4-9) is called the Lagrangian rate-distortion cost. The mode that minimises the Lagrangian rate-distortion cost for a particular λ≥0 (which satisfies the rate constraint in the constraint problem) is selected as the solution mode for the constrained problem. 4. Methodology

4.1 Objective video quality measurement

Objective video quality measurements are used to measure the video quality, typically in situations where fast (sometimes online) and repeatable measurements of the distortion or the difference between the video under test and a reference video are needed [7].

4.2 PSNR

The Peak Signal to Noise Ratio (PSNR) is the most commonly used objective measure of video quality. PSNR is measured as follows:

Where n is the bit depth and MSE is the Mean Squared Error between corresponding pixel values of the original image and the current image of the sequence under test. For M × N array of pixels, MSE is given by:

Where Po (i, j) denotes a pixel from the original image and Pi (i, j) denotes the corresponding pixel from the test image. The parameters ‘i’ and ‘j’ point to a position in the pixel arrays.

The MSE in itself can be a measure of distortion. However PSNR is preferred because the log scale provides a more realistic mapping to quality variations. Therefore, PSNR continues to be the most commonly used objective quality measure [5]. 5. Implementation


4

For the objectives to be achieved, software video simulation tool JSVM was used to implement and test the algorithms. There are plenty and different H.264/AVC reference software. JSVM was chosen for this research due to its flexibility of varying parameters.

JSVM codec is commonly used to test new algorithms in the video community. The use of this reference software enables realistic comparison of the performance of different algorithms developed by different researchers. The source code is mainly the same as the one used in the C programming language [8].

6. Results analysis In this part of the simulation, basic parameters such as frame rate of 30 Hz, number of frames 300 and group of pictures 16 were taken into consideration. Set of stiff video were used to evaluate the performance such as foreman, garden, football, flower, Claire and Carphone. The PSNR versus bitrates graph for various group of pictures were studied in difference circumstances. Below are different cases that were taken into consideration: Initially, spatial dimensioning is represented by QCIF and CIF, but was taken without additional progressive refinement (PR) slices. With additional PR, the transform coefficients are refined thus the improvement of the reconstructive pictures’ quality. The performance clearly proves that the PSNR varies with the quality.

Figure 2. Sequential scalable coding (Foreman)

In this case, several spatial resolution or bitrates are taken into consideration or provided by the encoded bitstream. The result shows that the PSNR is directly proportional to bitrates.

Figure 3. Single Layer coding

Based on Lagrangian Cost Function, if a video frame is to be sent on the outgoing link, it is first placed in the output buffer. Note that, for simplicity, we don’t consider the buffer limitations for the simulations in here.

If the outgoing link cannot accommodate all the video packets, it will first drop the additional enhancement PR slices one by one. If the link is still overloaded, the spatial enhancement layers are dropped next in the same spirit, i.e., scale out the enhancement layers completely sticking only to the base layer. The optimized SVC offers better quality than the unoptimized SVC one

Figure 4. Rate distortion optimization for scalable coding

Comparing to single-layer at higher bitrates, also, when the outgoing capacity Rout is larger than the required incoming rate, at 1670 kB/s, the RD-optimized single-layer coding and unoptimized coding perform the same.

This is obvious, as at higher bitrates, the network link will rarely overflow and very few or no video packets are lost. However, if the outgoing rate is very small, it can be seen that SVC strategy leads to good improvements in terms of reconstructed video quality. Table 1 shows the improvements obtained for individual video streams for the outgoing link Rout = 600 Kbit/s.

Figure 5. evaluation of SVC and SLC

Table 1: comparison of different video streams

Scalable Video Coding Single layer coding Sequences

Optimized (dB)

Unoptimized (dB)

Optimized (dB)

Unoptimized (dB)

Garden 45.0008 38.5645 42.5682 40.2658

Foreman 34.5545 35.2564 32.5654 30.1254


5

Football 37.2356 37.0052 36.2545 36.5485

Flower 40.3215 39.0235 37.5468 37.6256

Claire 36.2597 36.4566 31.2564 32.2564

Carphone 41.3255 38.4552 38.2545 39.2545

7. Recommendations Although the video coding standards exhibit acceptable quality-compression performance in many visual communication applications, further improvements are desired and more features need to be added, especially for some specific applications. The important considerations for video coding schemes to be used within future networks could be bases on Compression efficiency, robustness with respect to packet loss, adaptability to different available bandwidths and adaptability to memory and computational power for different clients. Several other communication and networking issues are also relevant, such as scalability, robustness, and interactivity. A network with a single active node was considered, in our simulations. This could be further enhanced to more practical situations with a hierarchy of many active network nodes and perform rate shaping at every node accordingly. Different values for the Lagrangian multiplier λ could be modeled for more stringent buffer conditions. A reasonable value for λ can be determined in maximizing the Lagrangian cost function, since λ is determined as a function of buffer fullness. The scalable video coding approach could be further extended to MCTF based scalable video codec which employs an open-loop architecture. 8. Conclusion The choice of a Scalable Video Coding framework in this context brings technical and economical advantages. Under this framework, network elements can adapt the video streams to the channel conditions and transport the adapted video streams to receivers with acceptable perceptual quality. The advantages of deploying such an adaptive framework are that it can achieve suitable QoS for video over wired and wireless networks, bandwidth efficiency and fairness in sharing resources [11].

The adaptive scalable video coding technology produces bitstreams decodable at different bitrates, requiring different computational power and channel bitrate. In addition, the bitstream is organized with a hierarchical syntax that enables users to easily extract only a subpart of the data contained in the bitstream and still being able to decode the original input video but at a reduced spatial resolution or frame rate. This process can be applied recursively, that is, once a new bit stream is extracted out of the original, it can undergo successive extractions corresponding to always lower resolutions.

References [1] C. S. Kannangara, I. E. G. Richardson, M Bystrom, J.

Solera, Y. Zhao, A. MacLennan & R. Cooney,

"Complexity Reduction of H.264 using Lagrange Optimization Methods," IEE VIE 2005, Glasgow, 4~6 April, 2005.

[2] H. Kim and Y. Altunbasak, "Low-complexity macroblock mode selection for H.264/AVC encoders," presented at International Conference on Image Processing, Singapore, 2004.

[3] K. P. Lim, "JVT -I020, Fast INTER Mode Selection." San Diego: ISO/IEC MPEG and ITU-T VCEG Joint Video Team, 2003.

[4] X. Li. Scalable video compression via over complete motion compensated wavelet coding. Signal Processing: Image Communication, special issue on subband/wavelet interframe video coding, 19:637—651, August 2004.

[5] S.-R. Kang, Y. Zhang, M. Dai, and D. Loguinov, \Multi-layer active queue management and congestion control for scalable video streaming," in Proc. IEEE ICDCS, Tokyo, Japan, Mar. 2004, pp. 768{777}.

[7] T. Oelbaum, V. Baroncini, T. K. Tan, and C. Fenimore, “Subjective quality assessment of the emerging AVC/H.264 video coding standard,” International Broadcasting Conference (IBC), Sept., 2004.

[7] R. Leung and D. Taubman. Impact of motion on the random access efficiency of scalable compressed video. Proc. IEEE Int. Conf. Image Processing, 3:169—172, September 2005.

[8] R. Leung and D. Taubman. Perceptual mappings for visual quality enhancement in scalable video compression. Proc. IEEE Int. Conf. Image Processing, 2:65—68, September 2005.

[9] R. Leung and D. Taubman. Transform and embedded coding techniques for maximum efficiency and random accessibility in 3-D scalable compression. IEEE Trans. Image Processing, 14(10):1632—1646, October 2005.

[10] R. Leung and D. Taubman. Minimizing the perceptual impact of visual distortion in scalable wavelet compressed video. Proc. IEEE Int. Conf. Image Processing, October 2006.

[11] R. Leung and D. Taubman. Perceptual optimization for scalable video compression based on visual masking principles. IEEE Trans. Circuits Syst. Video Technol., submitted in 2006.

[12] T. Wedi and Y. Kashiwagi, “Subjective quality evaluation of H.264/AVC FRExt for HD movie content,” Joint Video Team document JVT-L033, July, 2004.

[13] ISO/IEC JTC 1/SC 29/WG 11 (MPEG), “Report of the formal verification tests on AVC/H.264,” MPEG document N6231, Dec., 2003 (publicly available at http://www.chiariglione.org/mpeg/quality_tests.htm).

[14] T. Schierl, T. Stockhammer and T. Wiegand, "Mobile Video Transmission using Scalable Video Coding (SVC)," IEEE Trans. On Circuits and Systems for Video Technology, Special issue on Scalable Video Coding, scheduled June 2007.

[15] S. Wenger, Y.-K. Wang and T. Schierl, “Transport and Signaling of SVC in IP networks,” IEEE Transactions on Circuits and Systems for Video Technology, Special issue on Scalable Video Coding, scheduled for: March 2007.


6

Markov Based Mathematical Model of Blood Flow Pattern in Fetal Circulatory System

Sarwan Kumar 1 , Sneh Anand 2 , Amit Sengupta 3

1Dr B R Ambedkar National Institute of Techmology, Jalandhar -144011 ,Punjab,India

[email protected]

2Indian Institute of Techmology CBME Delhi, India

3Indian Institute of Techmology

CBME Delhi, India

Abstract: This paper presents a novel approach to estimate blood flow characteristics in the fetal circulatory system during pregnancy. We have developed a mathematical model of the fetal circulatory system by taking two nodes concept based on Markov model. As the oxygenated blood flows from mother side through placenta to fetus and deoxygenated blood from fetus to mother via umbilical cord. When it is simulated, the model shows how the oxygenated blood flows from placenta (one node) to the umbilicus (second node) and deoxygenated blood to placenta from the fetus. Also the same model is simulated at different conductivity path of the umbilical cord and available blood supply at placenta. Also it shows the effect of uterine contractions on the blood supply to the fetus. All simulations have been performed in the Lab VIEW environment at various conditions of vein and arteries.

Keywords: Markov model, placenta, umbilicus cord,

mathematical model, uterine contraction.

1. Introduction The baby develops in the uterus with a life support system for the fetus and is composed of umbilical cord, placenta and amniotic fluid. The placenta is a pancake shaped temporary organ which is attached to the uterus and is connected to fetus through the umbilical cord. The umbilical cord is the lifeline between the fetus and placenta. As soon it is formed it functions throughout pregnancy to protect the vessels that travel between the fetus and the placenta. The responsibility of the placenta is to act as a point of trade between the circulatory system of the mother and the baby. It is very important to know the relationship between concentration (quantity) of blood available at placenta and how quickly it passes to fetus through vein, the only path to carry good blood to fetus from mother and the waste products of the fetus are transferred to the mother’s blood through umbilical arteries. Therefore the umbilical cord is called the life line and it is through this cord that the placenta and the fetus are attached to each other. There are three blood vessels in the umbilical cord, two small arteries and a vein [10]. This cord can grow to a length of 50-60 cm which allows the baby to have enough space to move safely without damaging the placenta or the umbilical cord. The placental conductivity increases with the age of pregnancy [8]. Complete circulatory system is shown in figure 1.

During the time of pregnancy it can be found that the umbilical cord is in the form of a knot or at times this cord is wrapped around the body of the baby. This is a common phenomenon and there is no prevention for this. This does not pose any risk or threat to the baby or the mother. There may be some complications of the placenta due to pregnancy the most common of which is placenta previa [10]. In this the placenta is attached over or near the cervix. With the growth of the fetus there is pressure on the placenta and due to this reason there may be bleeding [11]. If this condition occurs there is need for medical care so that one can be ensured of a safe labor for the baby. Due to many complications there may be decrease in blood supply to fetus which leads to asphyxia and increase in heart rate [12]. Hence compromise of the fetal blood flow through the umbilical cord vessels can have serious deleterious effects on the health of the fetus and newborn. There for, it is necessary to know the blood profile in the fetus. This paper discuss about a novel mathematical model for the circulation of blood in the fetal circulatory system by taking two node concept based on Markov model [4] to know how the blood profile. The same model is simulated at various conductivities of the blood vessels and available blood at placenta. We have also demonstrated the effect of uterine contractions on the blood profile which would be useful to assist in developing a new bioelectric sensor for the evaluation actual blood flow time.

2. Markov Model A Markov model is a stochastic process whose dynamic behavior is such that its future development depends only on the present state space. In other words, the description of the present state fully captures all the information that could influence the future evolution of the process. Being a stochastic process means that all state transitions are probabilistic. At each step the system may change its state from the current state to another state (or remain in the same state) according to a probability distribution. The changes of state are called transitions, and the probabilities associated with various state-changes are called transition probabilities. In other words, the description of the present state fully captures all the information that could influence the future evolution of the process. In order to formulate a Markov model we must first define all the mutually


7

exclusive states of the system. The state of the system at t =0 are called the initial states (P0), and those representing a final or equilibrium state are final stage (P1). The set of Markov state equations describes the probabilistic transition from the initial to final states. The transition probabilities must obey the following two rules:

1 The probability of transition in time ∆t from one state to another is given by gain into ∆t.

2 The probabilities of more than one transition in time are infinitesimals of higher order and can be neglected.

3. Proposed Model Figure 1 presents the complete fetal circulatory system and its equivalent Markov chain in figure 2. The mostly problems related to node Ia are Intrauterine Growth Restriction (IUGR) and preeclampsia [5 ]. These are due to high blood pressure, diabetes, infection, kidney disease, heart or respiratory disease, alcohol, drugs and cigarette smoking ( figure 3) which may lead to fetal hypoxia, fetal death, low birth weight, placenta abruption (figure 4)[5]. The problems related to Umbilical Cord i.e. node II are two vessels, long cord, nuchal cord and short cord figure 5. The node Ib and node III are less significant in fetal circulation and are ignored. The modified nodes representation of the fetal circulation system and is equivalent signal flow graph is shown in figure 6. In term of mathematical model as described by the Markov Model [7], node I represents full of oxygen rich blood toward mother side and node II represents the fetus side. Umbilicus cord connects the two nodes.

There are two stages:

Stage 1: Placenta attached to the mother side, say node I, full of oxygen

Stage II: Umbilicus, the entering point to fetus, say node II

Let us P0(t) the quantity of the good blood at node 1

P1(t) the quantity of blood reaches at node II through vein gv the conductivity gain of the vein u1 the conductivity gain of the artery1 u2 the conductivity gain of the artery2 After ∆t the blood at node I and node II is given by

(1) P1(t+∆t) = P1(t) (1- u1 ∆t) + P1(t) (1- u2 ∆t) + P0(t) gv ∆t (2)

From equation 1 and 2

P0(t+∆t) - P0(t) = - P0(t) gv ∆t + P1(t)( u1 + u2) ∆t (3)

P0(t+∆t) - P0(t)/ ∆t = - P0(t) gv + P1(t)( u1 + u2) (4)

(5)

Similarly

(6) The final solution of equation’s 5 and 6 P0(s) = p0(s+u1 + u2) - p1(u1 + u2)/ (s+ gv)(s+ u1 + u2)- g v (u1 + u2) (7)

(8) And P1(s) = p1(s+ gv) + gv p1/(s+ gv)(s+ u1 + u2)- gv (u1 + u2) (9) P1(t) = (p1- p0)gv / (u1 + u2 )+ p1 (u1 + u2 - gv )/ (u1 + u2 )e- (u1 + u2) t (10)

Figure 1. Fetal Circulatory System

Figure 2. Fetal Circulatory System in nodes representation

(Node Ia Uterine Artery, Node Ib Placenta, Node II Umbilical Cord, Node III Fetal Heart)


8

Figure 3. Problems related to node Ia and their causes

Figure 4. Problems to fetus due to Intrauterine Growth Restriction (IUGR) and Preeclampsia

Figure 5. Problems related to node II (Umbilical Cord) 4. Simulation and analysis The software is designed in LabVIEW and simulated at different levels of P0, P1 ,gv ,u1 and u2. When equation 8 is simulated for various values of conductivities of vein ( gv) and two arteries (u1 , u2 ), we got the exponential curve. The response of the P0(t) is shown in figure 7 at gv unity or 100% conductivity. This indicates that the blood transfer from mother side placenta to fetus through vein having 100% conductivity at the start of the process. This indicates that the blood takes 4 second to reach to fetus. This exponential curve due to simulation is same as the current discharge through capacitor and register which has already been established in tissue impedance characterization [3],

[4]. Here tissue is represented by capacitor in parallel with resistance as shown in figure 8 [8 ] This time increases as the conductivity (gv ) decreases because of knot or some other reasons as shown in figure 9 where the time is approximately 10 seconds. This may leads to the child a number of dangerous effects (depression of the central nervous system, breathing paralysis, etc.).

Figure 6. Actual system and Equivalent Markov Signal Flow Graph

This will increase further if the contractions are more [1]. The quantity of the blood supply from the mother is highly affected the uterine contractions. Because the contractions will increase the intramyometrial pressure (120mmHg) compared to arterial pressure (85mmHg) [1]. The initial blood supply is less in this case and less amount of blood crossing the umbilical cord feeding fetus. This is the situation of less oxygen to the fetus, may lead to hypoxia. This is shown by our model as the output magnitude of P0 in term of available blood quantity. The output is shown in figure 10 for the 50% available blood. Maternal blood enters the placenta through the spiral arteries (the terminal branches of the uterine artery), which traverse the myometrium-muscular contractile layer of the uterus and flow into the intervelleous space. At this level the mother exchanges substances with the fetus through the "placental barrier." Anesthetics and analgesics are of low molecular weight, and are easily exchanged by diffusion as a result of the concentration gradient between the maternal and fetal compartments. Caldeyro-Barcia et.al have found that when the uterus is at rest, without contraction, the mother's arterial blood easily crosses the intervelleous space since the average arterial pressure is about 85 mmHg and the intramyometrial pressure external to the arteries is about 10 mmHg. During uterine contractions, however, the intramyometrial pressure rises to 120 mmHg, exceeding the arterial pressure which, under such conditions, is about 90 mmHg. The arteries therefore become temporarily occluded because of the external pressure, and the placenta becomes disconnected from the maternal circulation. [1] If the conductivity is reduces to 10%, it take approximately 50 second to reach its final stage ie fetal heart. This very dangerous stage for the fetus as lesser oxygen is going to


9

fetus extremely asphyxia. The fetus may die because of less blood or oxygen.

Figure 7. Output at gv unity with blood flow time of 5 seconds

Figure 8. Tissue’s Cell Membrane & its Electrical Equivalent

Figure 9. Output at gv 0.5 with blood flow time of 10 seconds

Figure 10. 50 % available blood during contractions and same amount to fetus

Figure 11. Deoxygenated blood flow curve, when both arteries are good

Figure 12. Deoxygenated blood flow curve, when one artery good

Figure 13. Blood flow time increases with decrease in length of the umbilical vein

When equation 10 is simulated at various levels of arteries path (u1 + u2 ). When both are working the response of this is shown in figure 11. It takes approximately 2 second to transfer the waste to placenta. This time increases as the conductivity decreases. The setting time is doubled when any one arteries is failed as shown in figure 12. This may increase the acidic or PH composition level in the fetus which can spoil umbilicus cord. The effect of length and diameter of vein have also been simulated. The reaching time also increases with the decrease in the length of the cord even the supply of the blood is 100% at placenta. The result of which is shown in figure 13. This indicates the blood is reached to fetus with longer time. When compared the result with the blood flow from model with the actual


10

flow of Doppler FVW taken from the paper [5], it shows the same blood flow pattern as actual. The figures 14(a) show the simulated blood flow while 14(b) the actual flow.

(a)

(b)

Figure 14. Comparison of the flow of blood between result from model and with actual flow, (a) Flow of blood response of Markov model (b) Actual blood flow: A frame extracted

from Doppler FVW [5] 5. Conclusion The blood flow timing between the placenta-fetus and fetus - placenta is given by the equation 8 and 10 respectively and simulated using LabVIEW software. The flow is exponential which shows that the umbilical cord structure (vein and arteries) acts as a capacitor in parallel to resistance. Time taken by the blood to reach fetus is increases as conductivity decreases. Also the time increases in case of lass quantity of blood is available due to uterine contractions, knot or any other reason. The simulated results show the larger settling time in case of short length. On the blood flow; it would be useful to assist in developing a sensor for the evaluation of conductivity of the umbilical cord and placenta during pregnancy for the well-being of fetus. We are developing a stand alone instrument for monitoring the various parameters of the fetal model. References [1] C. Hernandez Sande, G. Rodriguez-Izquierdo, and M.

Iglesias,” Intermittent Drug Administration During Labor and Protection of the Fetus, IEEE Transactions on Biomedical Engineering, pp 615-619, 1983.

[2] S M Sims, E Daniel and R E Garfield, “Improved Electrical Coupling in Uterine Smooth Muscle Is Associated with Increased Numbers of Gap Junction”, Journal of General of Physiology, pp-353-375, 1982.

[3] S Gandhi, D C Walker ,B.B. Brown and D. Anumba, “ Comparison of human uterine cervical electrical

impedance measurement derived using two tetrapolar probes of different sizes”, Biomedical Engineering ,pp 1-7,2006.

[4] R.J Halter,., A. Hartov,., J.A. Heaney, K.D. Paulsen,. A.R. Schned,., “Electrical Impedance Spectroscopy of the Human Prostate”, IEEE Transactions on Biomedical Engineering, pp 1321-1327 , 2007.

[5] A. Gaysen , S. K. Dua, A. Sengupta and Nagchoudhuri , “ Effect of Non-Linearity Doppler Waveforms Through Novel Model”, Biomedical Engineering Online, pp1-13,2003.

[6] A S Gordon ,, J Strauss and G A Misrahy, “ Electrical Impedance of Isolated Amnion”, Biophysical Journal, ,pp 855-865,2000.

[7] G. D. Clifford, F. Azuaje, P.E. McSharry , “Advanced Methods and Tool for ECG Data Analysis”, Artech House, pp 295-300 , 2006 .

[8] Guyton, Textbook of Medical Physiology, Eight Editions, 1991.

[9] Ross and Wilson, Anatomy and Physiology in Health and Illness, Tenth Edition , 2006.

[10] T. Erkinaro , “Fetal and Placental Haemodynamic Responses to Hypoxaemia , Maternal and Vasopressor Therapy in a Chronic Sheep Mode” l, Acta University , pp-1-96, 2006.

[11] J. C. Huhta , “ Fetal congestive heart failure” Seminars in Fetal & Neonatal Medicine 10, pp 542-552 , 2005.

[12] F. Kovacs, M. Torok, and I.Habermajer , “A Rule-Based Phonocardiographic Method for Long-Term Fetal Heart Rate Monitoring” , IEEE Transactions on Biomedical Engineering , pp 124-130 , 2000.

Authors Profile Sarwan Kumar received the BTech and MTech degrees in Electrical Engineering from Regional Engineering College Kurukshetra in 1992 and 1997, respectively. He is associate professor at National Institute of Technology Jalandhar. Now he is pursuing PhD from IIT Delhi, India under the guidance of professors Sneh Anand IIT Delhi and Dr. Amit Sengupta, , Consulting Obstetrician & Gynecologist (CHS), Mumbai.


11

Redevelopment of Graphical User Interface for IPMS Web Portal

Hao Shi

Victoria University, School of Engineering and Science Melbourne, Australia [email protected]

Abstract: IPMS, short for Industry Project Management System, is a web portal for industry project team management. IPMS is a very useful project managment tool to manage students, allocate projects, coordinate supervisors and liaise with industry sponsors. It has speeded up the process and allowed the stakeholders to focus on their key tasks. However the originally developed IPMS is no longer working after migrating to a new server. As a result, the manual management was brought back which was both times consuming and tedious for both project students and the course coordinator. This project aims to upgrade IPMS to PHP5.0 and re-develop the new GUI (Graphical User Interface) with enhanced system functionalities. In this paper, first the background information about IPMS is described. Then the newly developed GUI is presented and the usability test is conducted on the re-developed GUI. It is concluded that the newly developed GUI meets the user requirements and is better than the existing GUI.

Keywords: IPMS, Project Management Systems, GUI (Graphical User Interface), Industry Project, Web Portal.

1. Introduction Many final-year projects are offered at tertiary computing degree programs to provide project students team work and real world project experience under supervision of external project sponsor and academic staff [1]. However, managing software project teams is a complex task [2]. It should have occupied 20% time of the project coordinator but in the end it took more than 80% [3]. In order to reduce administrative load, many project management tools have been produced [4]. Some tools monitor full cycles of software engineering projects while others emphases more on aspects of the management projects [2].

SourceForge is the best known web portal, currently hosts over one hundred thousand projects and over a million users [5]. Open source tools such asDokuWiki, Trac, and Subversion can be integrated to provide a low-cost platform for student collaboration on team projects [6]. By consolidating project artifacts in a central location, the wiki software serves as both a repository for project information and a means of communication between team members and course instructors who may be working from different physical locations. Integrated version control helps students track changes in their documents and provides a safety net for recovery of information that has been previously deleted from project artifacts [6].

Even Moodle or Blackboard, dedicated eLearning tools [7] have incorporated group management tools which allow the

course designer to create groups and manage group activities besides course contents. Unfortunately there is no “one size fits all” solution in project management [8]. Many higher education institutions continue to build their own project management tools as they provide significant benefits to teaching and learning. More and more project teams in industry, academia, and the open source community are increasingly reliant on web-based project management portals [9].

The market for tools to improve software project management and software quality management is fast growing. It has been approved that vendors of software project and quality management tools can walk the talk by using quantitative data to manage the development project and process [10]. Recently, agile software development methods are popular because software should be developed in a short period. However, conventional project management techniques are often not adaptable to such new development methodologies. A new tool based on the communication model has been developed for agile software development [11], which allows to monitor product quality control and progress control. Some of these tools focus mainly on project management for teaching and learning [8] while others have full support for administrative tasks such as student registration, team formation, project confirmation, supervisor allocation and document management [4, 12, 13].

In this paper, it aims to upgrade the existing IPMS and re-develop a new GUI for IPMS.

2. Background IPMS (Industry Project Management System) is a “All-In-One” web portal. It was primarily developed to automate and streamline management of the final-year Industry Projects at Faculty of Health, Engineering and Science at Victoria University. IPMS prototype was initially developed based on the FDD methodology using Linux-Apache-MySQL-PHP, (LAMP) three-tier client-server architecture [Martin] as shown in Figure 1. PHP (Hypertext Preprocessor) is the program language which generates dynamic web GUI (graphic user interface). Apache is employed as a web server running under Linux operating system while MySQL is database management system for IPMS web pages and supports CMS (Contents Management System). LDAP (Lightweight Directory Access Protocol) is used for the user authentication [4, 11].


12

Figure 1. System topology [Martin]

Examples of the IPMS major menus are shown in Figure 2 and 3.

Figure 2. Student Menu [4]

3. Redevelopment of IPMS GUI In the 1st Semester 2009, one of industry project teams were assigned the task to upgrade the exiting IPMS and re-develop its GUI because it was no longer working after PHP was upgraded from version 4.0 to 5.0 on a new server. The project team consists of four final-year computer science students. They aim to maintain existing system functionality in the new system and improve the user interface and the logical flow of pages and add possible new functionality [14]. In the following subsections, the newly developed Admin menu and Student Menu are presented.

Figure 3. Admin Menu [4]

3.1 Top-down Module Design After the structured system analysis, the top-down design for IPMS module using Gane and Sarson graphical specification technique is shown in Figure 4.

3.2 Admin Menu Once a user is logged in as an admin user, the admin menu becomes available at the left side. The major change is that the menus are grouped into major functionality shown in shown Figure 5.


13

Figure 4. Top-down Module Design for IPMS [15]

Figure 5. Newly developed GUI for Admin Menu [15]

3.2.1 General Information Menu The general information menu consists of six submenus, namely About Industrial Project, Projects, Supervisors, Sponsors, Industrial Partners and FAQs as shown in Figure 6.

3.2.2 Administrator menu The Administrator Area contains the key menus such as Overview Information, Reports, Emails, User Administration, Project Administration, Team Administration, Database Access and Content Management as shown in 7.

Figure 6. General Information submenus

Figure 7. Administrator Menu

3.2.3 My Team menu My Team menu contains three submenus namely: RCM3001, RCM3002 and Assessment, each submenu contains several submenus as well as shown in Figure 7.

Add_New

Login Logout PHPMyAdmin

Check_Login

Update_Web_Content

Update

Register Database Web_Content User Email User_Details Login

IPMS Module

Project Report Former_Student

User_Report

Team_Report

Project_Report

Team

Team _Document

Details Delete

External

Back to the default home page

Contain general web pages that are meant for everyone to see

Contain forms/documents used to enable a supervisor to evaluate Team’s performance

Contain the assigned Team(s) to a supervisor with their contact details

Contain web pages which displays Team’s personal information

Contain forms/documents used to enable unit coordinator to able to modify user, project and team information. Logout from the system


14

(a) RCM3001 submenu

(b) Assessment submenu

Figure 8. My Team menu

3.3 Student Menu Once a user logins in as a student, the student menu displays as shown in Figure 8.

3.2.1 My Team menu This menu contains five submenus, i.e. Team Details, Create, Join, Upload Photo and Leave as shown in Figure 10.

4. Use Acceptance Test Many usability tests were carried out by the project coordinator during the course of IPMS re-development. Improvement and changes were made to enhance the GUI. In this paper, the usability test conducted is in the form of student user experience survey on the new developed GUI on the same aspects [12]:

Q1. Registration/Signup process Q2. Availability of other students to form a team Q3. Team formation Q4. Registration of an available project to a team or

project proposal Q5. Efficiency from registration to team formation Q6. Overall experience

The detailed results of the usability test are shown in Figure 11.

Figure 9. Student Menu [15]

Figure 10. My Team submenus

Q1. Registration/Signup process

0%

10%

20%

30%

40%

50%

Very Hard Hard Neutral Easy Very Easy

No.

of U

sers

Paper-BasedNEW GUI

(a)

Q2. Availability of other s tudents to form a team

0%

10%

20%

30%

40%

50%


No.

of U

sers

Paper-Based

New GUI

(b)

Contain general web pages that are meant for everyone to see

Contain web pages which displays Team’s personal information

Contain the team with their contact details

Contain forms/documents used to enable students to view Team’s performance criteria

Back to the default home page

Logout from the system


15

Q3. Team formation

0%

10%

20%

30%

40%

50%

60%


No.

of U

sers

Paper-BasedNEW GUI

(c)

0%

10%

20%

30%

40%

50%


No.

of U

sers

P aper-BasedIP M S Portal

Q4. Registration of an available project to a team or project proposal

(d)

Q5. Efficiency from registration to team formation

0%

10%

20%

30%

40%

50%

60%


No.

of U

sers

Paper-BasedNEW GUI

(e)

Q6. Overall experience

0%10%20%30%40%50%60%70%


No.

of U

sers

Paper-BasedNEW GUI

(f)

Figure 11. User experience survey

5. Conclusions IPMS has been upgraded to PHP5.0 after one-year development. The usability test has proved the new developed GUI efficient and user friendly. The new upgraded IPMS removes tedious manual process and provides smooth management functionalities for students, supervisors, and coordinator and industry sponsors. It is concluded that the newly developed IPMS meets the user requirements and is better than the previous version.

Acknowledgements The author would like to thank the project team, Riad El Tabbal (team leader), Leang Heng It, Jack Toke and Duncan Tu and the project supervisor, Associate Professor Xun Yi for their contributions in revitalising the IPMS GUI.

References [1] J. Ceddia and J. Sheard, “Evaluation of WIER – A Capstone

Project Management Tool”, Proceedings of the International Conference on Computers in Education (ICCE), pp. 777-781, 2002.

[2] J. L. Smith, S. A. Bohner, D. S. McCrickard, “Project Management for the 21st Century: Supporting Collaborative

Design through Risk Analysis”, Proceedings of 43rd ACM Southeast Conference, pp. 2-300- 2-305, 2005.

[3] G. Jones, “One Solution for Project Management”, Proceedings of SIGUCCS (The Special Interest Group on University and College Computing Servies) Fall Conference. pp. 65-69, 2001.

[4] H. Shi, “IPMS: A Web Portal for Industry Project Team Management”, International Journal of Communication, Vol. 7 No. 4, April 2007, pp. 111-116.

[5] Source-Forge, http://sourceforge.net [Accessed: Feb. 12, 2010]

[6] E. R. Haley, G. B. Collins, and D. J. Co, "The wonderful world of wiki benefits students and instructors", IEEE Potentials, Volume: 27, Issue: 2, pp. 21-26, 2008.

[7] Blackboard, http://blackboard.com [Accessed: Feb. 12, 2010] [8] Moodel, Open-source course management system

http://moodle.com [Accessed: Feb. 12, 2010] [9] A. N. Norita and P. A. Laplante, “ Software Project

Management Tools: Making a Practical Decision Using AHP”, Proceedings of the 30th Annual IEEE/NASA Software Engineering Workshop, 24-28, 2006.

[10] G. V. Seshagiri and S. Priya, "Walking the Talk: Building Quality into the Software Quality Management Tool", Proceedings of the Third International Conference On Quality Software (QSIC), pp. 67 – 74, 2003.

[11] N. Hanakawa and K. Okura, "A project management support tool using communication for agile software development", Proceedings of the 11th Asia-Pacific Software Engineering Conference (APSEC), pp. 316 - 323, 2004.

[12] R. Martin and H. Shi “Design and Implementation of IPMS Web Portal”, Proceedings of International Conference on Computers and Advanced Technology in Education (CATE), pp. 16-21, 2007.

[13] H. Shi, "Reshaping ICT Industry Projects - My Three-Year Experience", Proceedings of AusWIT06 Australian Women in IT Conference, 4-5 December, Adelaide, Australia, pp.36-46, 2006

[14] R. El Tabbal, L. H. It, J. Toke and D. Tu, “Redevelopment of Industry Project Management System”, Final-year Industry Project Design Report, School of Engineering and Science, Victoria University, November 2009.-

[15] R. El Tabbal, L. H. It, J. Toke and D. Tu, “Redevelopment of Industry Project Management System”, Software Design Document and User Manual, School of Engineering and Science, Victoria University, June 2009.

Author Profile

Hao Shi obtained her BE in Electronics Engineering from Shanghai Jiao Tong University, China and her PhD at University of Wollongong. She is now an Associate Professor and ICT Industry Project coordinator at School of Engineering and Science, Victoria

University. She has established Industry- Based Learning program at the School and won a number of Teaching and Leaning grants and awards. She is currently managing more than a dozen of ICT university scholarships with local industry partners via her grants from Victorian Government, Australia.


16

Analysis of Searching Techniques and Design of Improved Search Algorithm for Unstructured

Peer – to – Peer Networks

Dr. Yash Pal Singh1, Rakesh Rathi2, Jyoti Gajrani3 , Vinesh Jain4

1 Bundelkhand Institute of Engg. and Tech. Jhansi India

[email protected]

2 Govt.Engg.College,Ajmer Badliya Circle, NH 08, Ajmer [email protected]

3Govt.Engg.College,Ajmer

Badliya Circle, NH 08, Ajmer [email protected]

4Govt.Engg.College,Ajmer

Badliya Circle, NH 08, Ajmer [email protected]

Abstract: We study the performance of several search algorithms on unstructured peer-to-peer networks, both using classic search algorithms such as flooding and random walk, as well as a new hybrid algorithm proposed in this paper. This hybrid algorithm uses two level random walks for the adaptive probability search (APS). We compare the performance of the search algorithms on several graphs corresponding to common topologies proposed for peer-to- peer networks. In this paper it is found that Local Indices algorithm gives the average performance. Intelligent search and Routing Indices have higher bandwidth. Further work can be done on reducing the size of the query message subsequently it will reduce the bandwidth. APS is the efficient technique among all. Further it can be improved by proposed search algorithm which uses two-level k-walker random walk with APS instead of k-walker random walk. Advantages of two level walk will further reduce collision of nodes and can help in searching the distant nodes in the network. But it may slightly increase the response time. Keywords: peer-to-peer networks, adaptive probability search.

1. Introduction P2P network is a distributed network composed of a large

number of distributed, heterogeneous, autonomous, and highly dynamic peers in which participants share a part of their own resources such as processing power, storage capacity, software and files. The participant in the P2P network can act as a server and a client at the same time. P2P systems constitute highly dynamic networks of peers with complex topology. This topology creates an overlay network, which may be totally unrelated to the physical network that connects the different nodes (computers). P2P systems can be differentiated by the degree to which these overlay networks contain some structure or are created ad-hoc. Network structure here means the way in which the content of the network is located with respect to the network

topology. Unstructured, Loosely structured and Highly structured are various categories of P2P networks based on the control over data location and network topology. In this paper we are mainly concerned on comparative study of various available search Algorithms in unstructured P2P systems also it present the design of new proposed search algorithm for unstructured P2P systems.

2. Unstructured P2P systems

In unstructured networks, the placement of data (files) is completely unrelated to the overlay topology. Since there is no information about which nodes are likely to have the relevant files, searching essentially amounts to random search, in which various nodes are probed and asked if they have any files matching the query. These systems differ in the way in which they construct the overlay topology, and the way in which they distribute queries from node to node. The advantage of such systems is that they can easily accommodate a highly transient node population. The disadvantage is that it is hard to find the desired files without distributing queries widely. For this reason unstructured p2p systems are considered to be unscalable. However work is done towards increasing the scalability of unstructured systems. Napster, Gnutella, Kazaa, Morpheus[1] are various unstructured P2P systems.

3. Searching in unstructured Systems [4] Initially for the purpose of searching specific data item,

flooding which is basically BFS was used but it generates a large number of duplicate messages and also does not scale well so a number of alternative schemes have been proposed to address the above problem.


17

These works include iterative deepening, k-walker random walk, modified random BFS, two-level k-walker random walk, directed BFS, intelligent search, local indices based search, routing indices based search, attenuated bloom filter based search, adaptive probabilistic search, and dominating set based search.

Searching strategies in unstructured P2P systems are either blind search or informed search. In a blind search such as iterative deepening, no node has information about the location of the desired data. In an informed search such as routing indices, each node keeps some metadata about the data location. To restrict the total bandwidth consumption, data queries in unstructured P2P systems may be terminated prematurely before the desired existing data is found; therefore, the query may not return the desired data even if the data actually exists in the system. An unstructured P2P network can not offer bounded routing efficiency due to lack of structure.

The searching schemes in unstructured P2P systems can also be classified as deterministic or probabilistic. In a deterministic approach, the query forwarding is deterministic. In a probabilistic approach, the query forwarding is probabilistic, random, or is based on ranking.

Another way to categorize searching schemes in unstructured P2P systems is regular-grained or coarse-grained. In a regular-grained approach, all nodes participate in query forwarding. In a coarse-grained scheme, the query forwarding is performed by only a subset of nodes in the entire network.

4. Comparison of Existing Search Algorithms

Based on search method, Query forwarding, Message Overhead and node duplication various searching methods are compared as follows-

Algo-rithm

Search method

Query forward-ing

Message over-head

Node dupli-cation

Flooding BFS,

Blind

Broadcast High High

Iterative Deepning

BFS,

Blind

Broadcast High High

Local Indices

BFS,

Informed

Broadcast Medium Mediu

m

Directed BFS

BFS,

Informed

Partial Broadcast

Medium High

Intelligent Search

BFS,

Informed

subset of neighbor

Medium Mediu

m

Routing indices

BFS,

Informed

subset of neighbor

Medium Mediu

m

Std. random walk

BFS,

Blind

One neighbor

Low Low

k-walker random walk

BFS,

Blind

subset of neighbor

High High

2 Lvl k-walker random walk

BFS,

Blind

subset of neighbor

Low Low

APS BFS,

Informed

subset of neighbor

Medium

Medium

Based on scalability, response time (RT), success rate(SR)

and bandwidth various searching methods are compared as follows- Algorithm Search

method Query forward-ing

Message over-head

Node dupli-cation

Flooding No High Medium Low

Iterative Deepning

Yes High Medium Medium

Local Indices

Yes Medium Medium Medium

Directed BFS

Yes Medium Medium High

Intelligent Search

Yes Medium

Medium High

Routing indices

Yes Medium Medium High

Std. random walk

Yes High

Medium Low

k-walker random walk

Yes Medium Medium low

2 Lvl k-walker random walk

Yes Medium

Medium low

APS Yes Low High

Medium

Among those algorithms, Adaptive Probability Search

(APS) is the most efficient algorithm. APS is based on k-walker random walk and probabilistic (not random) forwarding. Another interesting algorithm is Two-Level Random Walk in which walkers are searching for an object in two levels. So it reduces the redundancy of nodes.

5. Adaptive Probability Search (APS) [6] In the Adaptive Probabilistic Search (APS) [6], it is

assumed that the storage of objects and their copies in the network follows a replication distribution. The number of query requests for each object follows a query distribution. The search process does not affect object placement and the P2P overlay topology.


18

The APS is based on k-walker random walk and probabilistic (not random) forwarding. The querying node simultaneously deploys k walkers. On receiving the query, each node looks up its local repository for the desired object. If the object is found, the walker stops successfully. Otherwise, the walker continues. The node forwards the query to the best neighbor that has the highest probability value. The probability values are computed based on the results of the past queries and are updated based on the result of the current query. The query processing continues until all k walkers terminate either successfully or fail (in which case the TTL limit is reached). To select neighbors probabilistically, each node keeps a local index about its neighbors. There is one index entry for each object which the node has requested or forwarded requests for through each neighbor. The value of an index entry for an object and a neighbor represents the relative probability of that neighbor being selected for forwarding a query for that object. The higher the index entry value the higher the probability. Initially, all index values are assigned the same value. Then, the index values are updated as follows. When the querying node forwards a query, it makes some guess about the success of all the walkers.

The guess is made based on the ratio of the successful walkers in the past. If it assumes that all walkers will succeed (optimistic approach), the querying node pro-actively increases the index values associated with the chosen neighbors and the queried object. Otherwise (pessimistic approach), the querying node proactively decreases the index values. Using the guess determined by the querying node, every node on the query path updates the index values similarly when forwarding the query.

Upon walker termination, if the walker is successful, there is nothing to be done in the optimistic approach. If the walker fails, index values relative to the requested object along the walker’s path must be corrected. Using information available inside the search message, the last node in the path sends an “update” message to the preceding node. This node, after receiving the update message, decreases its index value for the last node to reflect the failure. The update procedure continues along the reverse path towards the requester, with intermediate nodes decreasing their local index values relative to the next hops for that walker. Finally, the requester decreases its index value that relates to its neighbour for that walker. If we employ the pessimistic approach, this update procedure takes place after a walker succeeds, having nodes increase the index values along the walker’s path. There is nothing to be done when a walker fails.

Figure 1. Searching object using pessimistic approach of

APS with walkers.

Figure 1 shows an example of how the search process

works. Node A initiates a request for an object owned by node F using two walkers. Assume that all index values relative to this object are initially equal to 30 and the pessimistic approach is used. The paths of the two walkers are shown with thicker arrows. During the search, the index value for a chosen neighbour is reduced by 10. One walker with path (A,B,C,D) fails, while the second with path (A,E,F) finds the object. The update process is initiated for the successful walker on the reverse path (along the dotted arrows). First node E, then node A increase the value of their indices for their next hops (nodes F, E respectively) by 20 to indicate object discovery through that path. In a subsequent search for the same object, peer A will choose peer B with probability 2/9 (= 20 20+40+30), peer E with probability 4/9 and peer G with probability 3/9.

APS requires no message exchange on any dynamic

operation such as node arrivals or departures and object insertions or deletions. The nature of the indices makes the handling of these operations simple: If a node detects the arrival of a new neighbour, it will associate some initial index value with that neighbour when a search will take place.

If a neighbour disconnects from the network, the node removes the relative entries and stops considering it in future queries. No action is required after object updates, since indices are not related to file content. So, although our algorithm actively uses information, its maintenance cost on any of these events is zero, a major advantage over most current approaches.

5.1 Discussion on APS Each node stores a relative probability (an unsigned

integer) for each of its neighbours for each requested object. So for R such objects and N neighbours, O(R x N) space is needed.

For a typical network node, this amount of space is not a burden. On nodes with limited storage capacities, index values for objects not requested for some time can be erased. This can be achieved by assigning a time-to-expire value on each newly-created or updated index. Each search or update message carries path information, storing a maximum of TTL peer addresses. Alternatively, each node can associate the search and requester node IDs with the preceding peer in the path of the walker. Updates then follow the reverse path back to the requester. This information expires after a certain amount of time.The number of messages exchanged by APS method to terminate in the worst case will be (2 x k x TTL) where all walkers (k walkers) travel TTL hops and then invoke the update procedure, so the method has the same complexity with its random counterpart. The only extra messages that occur in APS are the update messages along the reverse path. This is where the two index update policies are used.

Along the paths of all k walkers, indices are updated so that better next hop choices are made with bigger probability. Learning feature includes both positive and negative feedback from the walkers in both update approaches. In the pessimistic approach, each node on the


19

walker’s path decreases the relative probability of its next hop for the requested object concurrently with the search. If the walker succeeds, the update procedure increases those index values by more than the subtracted amount (positive feedback). So, if the initial probability of a node for a certain object was P, it becomes bigger than P if the object was discovered through (or at) that node and smaller than P if the walker failed. Conversely, if many of our walkers hit their targets on average, the optimistic approach should be considered. This is the only invariant we require from our update process.

The learning process in the optimistic approach operates in an opposite fashion, Learning is important to achieve both high performance and discovery of newly inserted objects. Unlearning helps our search process adjusts to object deletions and node departures, redirecting the walkers elsewhere. All the nodes participating in the search get benefited from the process.

Besides standard resource-sharing in P2P systems, APS achieves the distribution of search knowledge over a large number of peers.

6. Performance of APS [6] The main metrics used to evaluate the performance of a

search algorithm are the success rate, the number of discovered objects (Hits per Query) and the number of messages produced.

Figure 2. Success rate vs. number deployed walkers for APS

and Random walk algorithms

Figure 3. Message production vs. number deployed walkers

for APS and Random walk algorithms

Figure 4. Hits per Query vs. number deployed walkers for

APS and Random walk algorithms

7. Two- Level Random Walk[7] It’s an efficient search algorithm which increases the total

number of nodes searched for a certain total number of search step, and reduces the redundancy or average number of times a particular node is searched. It works in the following manner. When a node wishes to send a query with a certain search key, it composes a search message and broadcasts it to k1 randomly selected neighbours. The message has an initial TTL1 = l1 hops. When an intermediate node receives this message, it checks the TTL1 timer. If the latter is still more than 0 then it decrements the timer by one, selects one random neighbour and forwards the message to it. This process continues until one of the nodes, say node E, receives the message with an expired TTL1 timer (i.e. TTL1 = 0). We call such a node an edge node. The message will then “explode” into k2 search messages forwarded from this node. Specifically, node E will compose a message with TTL1=0, and a second timer TTL2=l2. It will then randomly select k2 of its neighbours, excluding the one it just received the message from, and broadcast the message to them. Figure 1 shows an example illustrating this process. At level one, a source node sends k1 random messages to a set of k1 randomly selected nodes of its neighbours. This constitutes k1 threads (or random walks) which travel from the source node to the edge nodes (a node where TTL1 expires). Each of the k1 threads will then explode into k2 threads (with TTL2 = l2 ) at each of the k1 edge nodes. This algorithm reduces redundancy by decreasing the average number of times a node is searched. In the one-level k-walk algorithm k random threads are generated from the source and they are likely to have “thread collisions” (i.e. threads run into each other) especially near the source. This results in having redundant hits in the same nodes (nodes being searched multiple times). On the other hand, the two-level algorithm sends fewer threads from the source node which results in a smaller probability of thread collisions near the source. Each of the k1 threads will then explode into k2 threads once it is ”sufficiently” away from the source and the other threads. This way, the same number of search threads can be generated (k=k1*k2) but with a larger number of nodes searched and a smaller probability of redundant searches to the same nodes using the same number of total search steps.


20

8. Enhancing Performance of APS Proposed algorithm uses two-level random walk for the

existing APS algorithm[6] instead of k-walker random walk. Advantage of two-level walk[7] over one-level walk is that it increases the total number of nodes searched for a certain total number of search step, and reduces the redundancy or average number of times a particular node is searched. So collision of nodes can be further reduced and also distant objects can also be search efficiently. Two level walk will also help in further reducing message overhead. Only disadvantage will be increased in response time.

9. Algorithm of proposed Technique Assumptions

k1 = k2 = k = k3 – Number of walkers in each level ttlcount – counter for ttl value l1 = ttl2 = ttl - Time to live for each level level – variable for level number kcount – counter for k3 i.e. number of walkers Select a querying node Kcount = 0 level = 1

while (level <= 2) { while (kcount <= k3) { while (ttlcount <= ttl) { select a neighbouring node by applying APS and Process the node; if object is not found then increment ttlcount by one continue; else come out of the loop (exit); } increment kcount by one; } increment level by one; }

10. Conclusion In this research work, various searching techniques in

unstructured p2p networks are studied. Comparative study of these techniques is done. A new Search Technique is proposed which helps in further enhancing the performance of APS.

References

[1] Stephanos Androutsellis-Theotokis, ‘A Surver of Peer-To-Peer File Sharing Technologies’, White Paper, ELTRUN, Athens University of Economics and Business, Greece, 2002.

[2] V. Vishnumurthy and P. Francis. On heterogeneous overlay construction and random node selection in unstructured P2P networks. In Proc. IEEE Infocom, 2006.

[3] .MA. Jovanovic, Modelling large-scale peer-to-peer networks and a case study of gnutella. Master's thesis, Department of Electrical and Computer Engineering and Computer Science, University of Cincinnati, June 2000.

[4] .Xiuqi Li and Jie Wu, Searching Techniques in Peer-to-Peer Networks, Department of Computer Science and Engineering, Florida Atlantic University.

[5] V. Vishnumurthy and P. Francis. A comparison of structured and unstructured P2P approaches to heterogeneous random peer selection. In Proc. Usenix Annual Technical Conference, 2007.

[6] D. Tsoumakos and N. Roussopoulos. Adaptive Probabilistic Search (APS) for Peer-to-Peer Networks. Technical Report CS-TR-4451, Un. of Maryland, 2003.

[7] Imad Jawhar and Jie Wu, A Two-Level Random Walk Search Protocol for Peer-to-Peer Networks, Department of Computer Science and Engineering, Florida Atlantic University.

[8] Beverly Yang Hector Garcia Molina, Improving Search in Peer-to-Peer networks, Computer Science Department, Stanford University.


21

Steganography Security for Copyright Protection of Digital Images Using DWT

K.T.Talele1, Dr.S.T.Gandhe2, Dr.A.G.Keskar3

1 Electronics Engineering Dept.,Sardar patel Institute of Technology,

Andheri(w)Mumbai, India [email protected]

2 Electronics Engineering Dept.,Sardar patel Institute of Technology,

Andheri(w)Mumbai, India [email protected]

3Electronics Engineering Dept , Visvesvaraya National Institute of Technology,

Nagpur,India [email protected]

Abstract: The proposed system combines cryptography and steganography for copyright protection of digital images using DWT. The proposed algorithms tested on various attacks such as median, wavelet compression, fading and resizing by comparing different performance parameters such as mean square error, peak signal to noise ratio, correlation coefficient and the results are very encouraging. The sensitivity is least observed in DWT method where the watermark maintains a fair level of resistance to noise and other attacks. The proposed system can be used for enhanced copyright protection, detection of misappropriated images; detect alternation of images stored in a digital library.

Keywords: Cryptography, encryption, Decryption, Steganography.

1. Introduction Security is one of the major concerns in today’s age. Unlike the past, most of the transactions between people take place over the internet. But internet itself is not a secure medium. So, when it comes to sending highly important documents over the internet, an extra precaution has to be taken. In other cases, authenticity of digital data is a big concern. With the widespread usage of digital media, demand for copyright protection has increased manifold as it is evidently seen in the audio records industry. The extra precaution for copyrighting digital media is required here as well. One of the ways to take this extra precaution is to use Steganography. Steganography helps to hide the content of interest which is to be protected, inside any image, audio or video file. To further ensure that interception of content does not happen, the content can be encrypted using one the popular Cryptographic algorithms[1][2]. Fragile Watermarking is used in the case where tamper detection and authenticity have a higher priority whereas Robust Watermarking deals with copyright protection[3][4] [5][6]. The proposed system hides a logo in images using DWT . The watermark should be imperceptible to anyone and sensitive to any kind of tampering done on the image under consideration. The system is compared for various algorithms for embedding the logo. The algorithms are compared on the basis of Mean Square Error (MSE), Peak

Signal to Noise Ratio (PSNR) and Correlation Coefficient (CC) values of the extracted logo for different attacks[7].

2. Proposed System The block diagram for proposed system is as shown in figure 1.First the logo is encrypted and then it is inserted in a given image using DWT and the logo is extracted and then it is decrypted to get original logo.

Figure 1. Block Diagram of Proposed System

3. Algorithm

3.1 Cryptography Algorithm The encryption algorithm works on the approach of swapping pixel values of randomly generated 128 locations in the row of every logo. It is important that the set of 128 locations so generated are done with the help of a password and that they are all unique. The steps of the algorithm are as follows:

a) Take the input logo. b) Ask the user to enter a 8 bit key. c) Generate 8 random vectors of size 1X128.


22

d) Specify the ‘state’ of the random number generator by giving the ASCII value of each character in the key for every random vector. This will generate 8 random vectors.

e) Generation of Input vector: Ø The outer loop controls the column

traversal of the watermark logo. Ø The middle loop controls the selection

of random vectors previously generated. Ø The inner loop controls the row

traversal of the watermark logo. Ø Inside the innermost loop, we select a

random vector based on the value of the middle loop. With every turn of the innermost loop we take two consecutive values, r(1,k,j) and r(1,k+1,j) from this random vector and swap the corresponding location values from the watermark image in the same column and store it in another array called ‘encrypted’ at the same position.

f) Every time the middle loop finishes, the random vectors are considered corresponding to the first character of the password.

g) This cycle continues till the last column is covered. h) Thus the watermark image is encrypted into a new

image file ‘encrypted.bmp’.

3.2 Proposed Steganography in color images Algorithm

a) Consider any color image having size 512 X 512 as

a host image. If size of host image is not 512 X 512 then make it 512 X 512.

b) Split the image into three planes viz. Red, Green and Blue.

c) Decompose the host image (all three planes) by using discrete wavelet transform. Store the first level coefficients i.e. LL1, LH1, HL1, HH1 as first level watermark key coefficients of host image.[8][9]

d) Approximation coefficient of first level is LL1 which is further decomposed into new coefficients i.e. LL2, LH2, HL2, HH2 as second level watermark key coefficients of host image.

e) Consider the gray scale image having size 128 X 128 as a logo to be hidden. If size of watermark logo is not 128 X 128 then make it 128 X 128.

f) Decompose the watermark logo by using discrete wavelet transform. Store the first level approximation coefficients i.e. LL1, LH1, HL1, HH1 as first level watermark key coefficients of watermark logo.

g) Insert coefficients of LH2 part of host image by LL1 part pixel by pixel.

h) Perform the two level inverse discrete wavelet transform of host image (all three planes) by using approximation coefficients of three planes of host image.

i) Find the mean square error (MSE) and peak signal to noise ratio (PSNR) and the correlation coefficient (CC) between the original host image and invisible watermark image by using the related formulae as these are the important performance parameters.

3.3 Performance Parameters

3.3.1 Peak Signal to Noise Ratio(PSNR) and Mean Square Error(MSE).

The imperceptibility of a watermark is measured by the watermarked image quality in terms of Peak-Signal-to-Noise Ratio (PSNR) (in dB). Most common difference measure between tow images is the mean square error. The mean square error measure is popular because it correlates reasonably with subjective visual tests and it is mathematically tractable. Consider a discrete image A(m, n) for m=1,2,……M and n=1,2,……N, which is regarded as a reference image. Consider a second image Â(m, n), of the same spatial dimension as A(m, n), that is to be compared to the reference image. Under the assumption that A(m, n) and Ã(m, n) represent samples of a stochastic process, MSE is given as

Where E (·) is the expectation operator. The normalized Mean Square Error is given as

Normalized mean square error for deterministic image

arrays is defined as

Image error measures are often expressed as signal-to-noise ratio,

We use PSNR to determine the difference between original image A (m, n) and the watermarked image Ã (m, n). The value of mean square error should be minimum and the value of peak signal to noise ratio should be as maximum as possible.


23

3.3.2 Correlation Coefficient (CC) The robustness performance of watermark extraction is evaluated by normalized correlation coefficient, r, of the extracted watermark A and the original watermark B.

Where A and B respectively, the normalized original and watermark image by subtracting its corresponding means value. The magnitude range of r is [0, 1], and the unity holds if the extracted image perfectly matches the original one. The correlation coefficient is used to compare original image and the watermarked image, and also for comparing original watermark and the retrieved watermark.

4. Results and Discussion The algorithms with various attacks are implemented using MATLAB. The results of various algorithms are shown through figure 2 to figure 19. 4.1 Cryptography

(a) (b) (c)

Figure 2: (a) Original Logo (b) Encrypted Logo (c) Decrypted Logo

Figure 2 (a) shows original logo and figure 2(b) shows encrypted logo and figure 2(c)shows decrypted logo.

4.2 Steganography 4.2.1 Steganography in images using DWT

(a) (b) Figure 3: (a) Original Image (b) Watermarked Logo (to be hidden)

(a) (b)

Figure 4: (a) Original Image (b) Extracted Logo

Figure 3(a) shows original logo and figure 3(b) shows watermark logo .Figure 4(a) shows watermarked image and figure 4(b) shows extracted logo.

4.3 Comparative Study Of Different Watermarking Algorithms for different original images

The algorithm of insertion is applied to five different input images as shown in figure 5 through figure 9 and compared the result using different performance parameters. The comparison is as shown in table 1.. We can insert secret information in these images for copyright protection.

(a) (b)

Figure 5: (a) Original Image (b) Watermarked Image for lena image.

(a) (b)

Figure 6: (a) Original Image (b) Watermarked Image for medical image

(a) (b)

Figure 7: (a) Original Image (b) Watermarked Image for satellite image.

(a) (b)

Figure 8: (a) Original Image (b) Watermarked Image for satellite Scene.


24

(a) (b)

Figure 9: (a) Original Image (b) Watermarked Image for text image.

Table 1: PSNR, MSE, and CC for original image and watermark image for Invisible Watermarking for five

different images

The value of PSNR is sufficiently high, MSE is very low and CC is nearly equal to 1.So this algorithm has created minimum disturbance to host image and perceptually both the images are alike.

(a)

(b)

(c) Figure 10: Graph of (a) PSNR, (b) MSE, (c) CC for steganography in images using DWT Graphically these values are shown in figure 10.

4.4 Attacks The algorithm is tested for various attacks[7] such as median filter, wavelet compression, fading, noise, resizing etc.

4.4.1 Median Filter The figure 11 through figure 14 shows the algorithm is tested for median filter under four mask sizes 3X3, 5X5, 7X7, 9X9. For each of these cases peak signal to noise ratio, mean square error and correlation coefficient are calculated and are as shown in figure 15.

(a) (b) Figure 11: (a) Median filtered Watermark Image with mask

size 3X3, (b) Extracted Watermarked Logo

(a) (b) Figure 12: (a) Median filtered Watermark Image with mask

size 5X5, (b) Extracted Watermarked Logo


25

(a) (b) Figure 13: (a) Median filtered Watermarked Image with mask size 7X7, (b) Extracted Watermark Logo

(a) (b)

Figure 14: (a) Median filtered Watermarked Image with mask size 9X9, (b) Extracted Watermark Logo

(a)

(b)

(c) Figure 15: Graph of (a) PSNR, (b) MSE, (c) CC for watermarked image for different mask size of median filter

4.4.2 Wavelet Compression.

The algorithm is tested for wavelet compression and is as shown in figure 16. The performance parameters are MSE=0.0060, PSNR=46.02619, CC=0.9907

(a) (b) Figure 16 (a) Wavelet compressed Watermarked image (b)

Extracted Watermark Logo

4.4.2 Fading The algorithm is tested for fading where each pixel value of image is increased by 50 and extracted watermark logo and is as shown in figure 17. The performance parameters are MSE=0.0561, PSNR=36.5775, CC=0.9954

(a) (b)

Figure 17: (a) Faded Watermarked Image (Original+50), (b) Extracted Watermark Logo

4.4.3 Noise

(a) (b)

Figure 18: (a) Noise added Watermarked image, (b) Extracted Watermark Logo

The algorithm is tested for noise and is as shown in figure 18. The performance parameters are MSE=0.0322 PSNR=38.9902 CC=0.8103

4.4.4 Resizing The algorithm is tested for resizing where watermark image resized by a scaling factor of 2.and is as shown in figure 19.The performance parameters are MSE=0.0028 PSNR=49.5556 CC=0.9972.


26

(a) (b) Figure 19: (a) Resized Watermarked Image, (b) Extracted

Watermark Logo

5. Conclusion Thus we have developed a system which successfully embeds a logo imperceptibly in a given cover image. The logo is sensitive to tampering and compression. On the other hand, the sensitivity is least observed in DWT method where the watermark maintains a fair level of resistance to noise and other attacks. In order to further prevent the watermark logo from detection, we encrypt it before embedding it in an image. The algorithm used is based on swapping the pixel values of the watermark logo by using password generated random vectors. The system is not immune to compression techniques. Whenever the watermarked is stored in compressed format, a considerable loss of information is observed. The system can be extended to possible counter the compression effect. Extracted watermark face image can be recognized by using automated intelligent face detection algorithm[11].The proposed system can be used for enhanced copyright protection, detection of misappropriated images, detect alternation of images stored in a digital library.

References [1]W.Diffie and M.E.Hellman, “New Directions in

Cryptography”, IEEE trans. On Information Theory, Vol.IT-22, No.6, Nov.1976.

[2] B.M.Macq, J.J.Quisquater, “Cryptography for Digital TV Broadcasting”, Proc. of the IEEE, Vol.83, No.6, Jun1995, pp 944-957.

[3] J. M. Acken, “How Watermarking Value to Digital Content”, Comm. of ACM, July 1998, Vol 41, No.7, pp 75-77.

[4] Yongjian Hu, Sam Kwong, Jiwu Huang, “An Algorithm for Removable Visible Watermarking”; IEEE transactions on circuits and systems for video technology, vol. 16, no. 1, January 2006.

[5] Zhang Fan, Zhang Hongbin, “Capacity and Reliability of Digital Watermarking”, International Conference on the Business of Electronic Product Reliability and Liability 2004.

[6] Fan Zhang, Hongbin Zhang, “Digital Watermarking Capacity and Reliability”, Proceedings of the IEEE International Conference on E-Commerce Technology 2004.

[7] Chun-Hsiang Huang and Ja-Ling Wu, “Attacking Visible Watermarking Schemes”, IEEE transactions on multimedia, vol. 6, no. 1, February 2004

[8] S.T.Gandhe and K.T.Talele and Dr. A.G. Keskar “ Face Recognition Using DWT+PCA” International

conference INDICON07 organized by IEEE India council at Bangalore on 6th,7th,8th September 2007.

[9] S.T.Gandhe, K.T.Talele, A.G.Keskar “Intelligent Face Recognition: A Comparative Study” ICGST’s Graphics Vision & Image Processing Journal, Vol.7,Issue 2,2007, page no.53-60.

Authors Profile

Dr. A.G.Keskar is a Professor in Electronics Engg. Dept,Visvesvaraya National Institute of Technology, Nagpur. He is a senior member of IEEE. He has published 10 Journal papers and published 25 papers in International Conferences. His area of interest includes Fuzzy logic, Embedded System and Machine Vision.

Dr.S.T.Gandhe is Professor in Electronics Engg Department, S. P. Institute of Technology, Mumbai, India.. He is a member of IEEE. His area of interest includes Image Processing, Pattern Recognition and Robotics. He has published 12 papers in National Conferences and 10 papers in International conference and 3 papers in international journal.

K.T.Talele is aAssistant Professor in Electronics Engg Dept, S. P. Institute of Technology, Mumbai., India He is a member of IEEE. His area of interest includes DSP, Image Processing and Multimedia Comm. He has published twenty papers in National Conferences and three papers in International journal and 14 papers in international conference


27

Design and Implementation of A GUI Based on Neural Networks for the Currency Verification

1Ajay Goel , 2O.P.Sahu, 3Rupesh Gupta and 4Sheifali Gupta

1Department of CSE, Singhania University, Rajasthan, India, [email protected]

2Department of ECE, N.I.T. Kurukshetra, India,

[email protected]

3Department of ME, Singhania University, Rajasthan, India, [email protected]

4Department of ECE, Singhania University, Rajasthan, India,

[email protected]

Abstract: The technological development in the era of image processing and machine vision has two faces. One face is to help the society by automation and the other side has serious implications on the society like cyber crimes e.g. web hacking, cracking, etc. One of the emerging crimes is preparing fake legal documents in now days. These documents have social values, like a degree certificates certifies the educational qualification of a person. The legal documents contain lots of symbols like kinegrams, hologram, watermark etc. by which we can verify the authenticity of these documents. Digital watermarking emerged as a tool for protecting the multimedia data from copyright infringement. In this paper an attempt has been made to verify the legal document on the basis of watermark. In this work the correlation mapping with neural network is used for extracting the watermark to verify the legal documents. This technique gives elevated accuracy rate with fewer times to extract watermark. This method can be implemented also in additional applications like stamp verification, currency verification etc.

Keywords: Watermarking, Multilayered-Network, Certifying, Epochs.

1. Introduction With the increasing use of internet and effortless copying, tempering and distribution of digital data, copyright protection [1] for multimedia data has become an important issue. There are lots of symbols present on the printed document for their identification but in this work watermark has been chosen for the verification. Basically legal documents can be verified by two methods: first-line inspection methods and second-line inspection methods .First-Line Inspection Methods are Watermarks, Ultraviolet Fluorescence, Intaglio Printing are further divided in Micro text & Holograms and Kinegrams and second is Second-Line Inspection Methods Isocheck / Isograms. Recent public literatures show that some researchers have tried to apply watermarking into printing system. In geometric transform. Pun [5] has devised a watermarking algorithm robust to printing and scanning. The PhotoCheck software developed in AlpVision Company by Kutter [6] is mainly fo-cused on authentication detection of passports. As a passport belongs to the owner with his photo, this belongs to

a content-based watermarking technique. When the photo is changed, the image with the watermark is of course lost and this just requires that the watermark hidden in the owner’s passport is robust to one cycle of print and scan. Considering the special characteristic of FFT on rotation, scaling and cropping, Lin [7][8] has carried out the research on fragile watermarking rather early and obtained many useful conclusions on the distortion brought by print and scan. Re-searchers in China [9] began to hide some information in printing materials, using the function offered by PhotoShop. All these are focused on the watermark robust to one cycle of print and scan.

2. Basic Concept 1. Since water mark making requires highly efficient

technique and the water mark can be seen only by its shadow, the water mark can be effective key to certify the currency note.

2. In certifying the currency note, since normally using currency note is folded, sometimes noise occurred, it needs feedback learning of water mark of used currency note. The back propagation neural network is suitable to certify the water mark, because it can design many layers for many nodes network, that it is used to recognize the complicate pattern [3].

They consider correlation as the basis for finding matches of a sub- image w(x,y) of JxK within an image f(x,y) size MxN, we assume that J≤M and K≤N. They prepare the template of each type of note then apply correlation on each stored note with on which we are testing. Zero value of correlation coefficient gives the location of the watermark[10].

3. Certifying To certify the watermark it is inputted to back-propagation neural network. Result of neural network is used to certify the currency note. First neural network must be trained by sending the ideal watermark to it. Size of input is sent to neural network about 4225 nodes.


28

4. Trainable Multilayered- network

BP method is given by eq(5.2) as [6]

)( tkWxW ijjiij −∆+=∆ αηδ (1.1) Where ijW is the weight connecting an output of unit j into an input of unit i, η is the step size, a is a momentum coefficient, and xj is an input signal from unit j. The quantity iδ is an error term, computed as

−

= ∑ mmiWiOOitiiO

i δδ ')('

(1.2)

Where ti is a desired signal to unit I, Oi is the actual output of unit i, and O’i is the derivative of O’i. Weight is adjusted according to what is the target. When there is much difference between input and target then neural network needs more training. When epochs are increased neural network gives best response as the number of epochs increased training is better, although with increasing number of time taken to train the network is more but target is achieved with less mean squared error [4].

5. Methodology In this methodology first the searching of watermark is done for that a document is split into blocks. Block size of each block is equal to the size of ideal watermark. After splitting each block is stored in different variable. After splitting document into blocks further process is that to correlate each blocks with ideal watermark. This correlation will give the correlation coefficient. Block which contain watermark gives the highest correlation coefficient. Now we extracted the block which gives us highest correlation coefficient and give it to neural network. Correlation coefficient[3] is given by the equation (1.3)

∑∑ −−= ),(),(),( tysxwyxftsc (1.3)

Where x, y are co-ordinates of selected block and s=0, 1, 2…. M, t= 0,1,2,……N Where M×N is size of ideal watermark. Each pixel of the selected block is matched with each corresponding pixel of ideal watermark block. Difference between these is calculated called correlation coefficient. Correlation coefficient of each block has been calculated with the same procedure. Now select the block which contains the highest correlation coefficient. The system software flowchart is shown in figure 1 can be described; the location of the water mark is detected[7].

Figure 1. Flowchart of a system

The edge information from the shadow of water mark is derived from shining image, and it is inputted to neural network for certifying. The edge information all of currency note is inputted to neural network for checking.

6. The Proposed Scheme We are applied an algorithm which tries to find the watermark more effectively. First step to break the image into different pieces having size of each piece is equal to ideal watermark size. Then it finds the correlation among each piece with ideal watermark. Correlation process gives correlation coefficients then the method picks up the piece which gave greatest correlation coefficient, because this piece has most probably of containing watermark. The piece found in the previous step has been given to the neural network for verification. We trained the neural network with ideal watermark as a target. This is implemented in matlab 7.0 and vb.net. GUI has been created in vb.net and main implementation has been done in matlab 7.0. Then GUI is linked with matlab7.0 with the help of M-files.

Figure 2. Main form of GUI, “Document verification system”

In fig 2 photo of the currency note 1000 has been shown. After acquiring image the neural network has been trained. Following figure 3 shows the training of neural network.

Figure 3. TrainingoOf Neural Network


29

After clicking ok main form displayed again then we have to verify it. By selecting verify from the menu bar will start process of verification. Following graph in figure 4 shows the watermark accuracy present on the document.

Figure 4. Accuracy of two watermark present on the

document.

7. Result The note is divided into two categories one is called training set other is called testing set. After inputting the note’s image of currency notes, the location of watermark has been detected by correlation mapping by splitting image into blocks then finding correlation of each block. Then block is given to neural network for certification. Neural network is trained through training set. After the goal is accomplished we test the neural network by tested data. In this neural network Mean squared error has been used. So training graph has been plotted target versus input. So the neural network has been trained with 1000 epochs.

Figure 5. Training neural network with 1000 epochs

This bar chart is plotted on scale 1. First watermark shows the accuracy of 99% and second watermark shows the accuracy of 100% which are above threshold so we can say that the currency note is real. This accuracy is checked on the basis of percentage matching with real watermarks because neural network has been trained with the real watermark and gives mean squared error when tested. Two neural network for two watermark has been created differently. If accuracy is below threshold then the note is rejected.

Figure 6. Bar chart showing percentage accuracy of two watermarks.

We have implemented our technique on the Indian currency and Indian postage stamp, but their technique is implemented on Thai currency. However output of this technique is also different it did not show the accuracy of watermark. This technique searches the watermark into whole image while our technique will split the image into blocks and apply correlation on each block with ideal watermark, which gives us a correlation coefficient. The value of correlation coefficient gives us an idea of similarity between two images. This technique takes shorter time to find the watermark in the note.

Reference [1] Ingermar J. Cox, Matthew L. Miller, and Jeffrey

A.Bloom, Digital Watermarking, Morgan KaufmannPublishers, 2002

[2] Francisco J. Gonzalez-Serrano, Harold. Y. Molina- Bulla, and Juan J. Murillo- Fuentes,” Independent component analysis applied to digital image watermarking,” International Conference on Acoustic, Speech and Signal Processing (ICASSP), vol. 3, pp. 1997-2000, May 2001.

[3] Dan Yu, Farook Sattar, and Kai-Kuang Ma, “Watermark detection and extraction using independent component analysis method,” EURASIP Journal on Applied Signal Processing, vol. 1, pp. 92–104, 2002.

[4] Minfen Shen, Xinjung Zhang, and Lisha Sun, P. J. Beadle, F. H. Y. Chan, “A method for digital image watermarking using ICA,” 4th International Symposium on Independent Component Analysis and Blind Signal Separation (ICA 2003), Nara, Japan, April 2003, pp. 209-214.

[5] Ju Liu , Xingang Zhang, Jiande Sun, and Miguel Angel Lagunas, “A digital watermarking scheme based on ICA detection,” 4th International Symposium on Independent Component Analysis and Blind Signal Separation, (ICA 2003), Nara, Japan, April 2003, pp. 215-220.

[6] Stephane Bounkong, Boremi Toch, David Saad, and David Lowe, “ICA for watermarking digital images,” Journal of Machine Learning Research 4, pp. 1471-1498, 2003.

[7] Viet Thang Nguyen and Jagdish Chandra Patra, “Digital image watermarking using independent component analysis,” PCM 2004, Lecture Notes in Computer Science 3333, pp. 364-371, Springer-Verlag, 2004.

[8] Thai Duy Hien, Zensho Nakao, and Yen-Wei Chen, “Robust multi-logo watermarking by RDWT and ICA”, Signal Processing, Elsevier, vol. 86, pp. 2981-2993, 2006.

[9] Aapo Hyvarinen, “Survey on Independent Component Analysis”, Neural Computing Surveys,vol. 2, pp. 94-128, 1999.

[10] Hyvarinen, Karhunen, and Oja, “Introduction,” Chapter 1 in Independent Component Analysis,John Wiley, pp. 1-12, 2001.


30

An Approach to Protection at Peer-To-Peer Networks

*Dr.G.Srinivasa Rao, *Dr.G. Appa Rao, *S. Venkata Lakshmi, *D. Veerabhadra Rao, **B. Venkateswar Reddy,*P. Venkateswara Rao, *K. Sanjeevulu.

*GITAM University, **CVSR Engineering College [email protected]

Abstract: Open networks are often insecure and provide an opportunity for viruses and DDOS activities to spread. To make such networks more resilient against these kind of threats, we propose the use of a peer-to-peer architecture whereby each peer is responsible for: (a) detecting whether a virus or worm is uncontrollably propagating through the network resulting in an epidemic; (b) automatically dispatching warnings and information to other peers of a security-focused group; and (c) taking specific precautions for protecting their host by automatically hardening their security measures during the epidemic. This can lead to auto-adaptive secure operating systems that automatically change the trust level of the services they provide. We demonstrate our approach through a prototype application based on the JXTA peer-to-peer infrastructure. Keywords: Peer-to-peer, Antivirus, Intrusion Detection, JXTA 1. Introduction

The rapid evolution of the Internet, coupled with the reduction in the cost of hardware, have brought forth very significant changes in the way personal computers are used. Nowadays, the penetration of the Internet is wide, at least in the developed world, and high percentage of connectivity is handled through broadband technologies such as DSL, cable modems, satellite links and even 3G mobile networks. Many companies have permanent connections to the Internet through leased lines and optical fibers, and many home users through the aforementioned broadband connections. If one also takes into account the significant development of wireless networking technologies (such as Wireless LAN, HyperLAN), the immediate result is an almost universal connection of most users on a 24-hour basis. Although the potential benefits arising from these developments are various and important, so are the dangers that follow from the possibility of malicious abuse of this technology. The proliferation of viruses and worms, as well as the installation of Trojan horses on a large number of computers aiming at Denial of Service (DoS) attacks against large servers, constitute one of the major current security problems. This is due to the extent to which critical infrastructures and operations such as hospitals, airports, power plants, aqueducts etc. are based on networked software-intensive systems. The measures taken for protection against such threats include [45] the use of firewalls, anti-virus software and intrusion detection

systems (IDS). Considerable importance is also placed on the topology of the network being protected [43], as well as to its fault tolerance to ensure that its operation will continue even if a part of it is damaged. A significant increase in the spread of viruses, worms and Trojan horses over the Internet has been observed in the recent years. Recent evidence shows that older boot sector viruses, as well as viruses transmitted over floppy disks no longer constitute a considerable threat [12]. At the same time, though, modern viruses have become more dangerous, employing complex mutation, stealth and polymorphism techniques [37] to avoid detection by anti-virus software and intrusion detection systems. These techniques are particularly advanced and, combined with the fact that antivirus software is often not properly updated with the latest virus definitions, can lead to uncontrollable situations. In the last two years it has been proven both theoretically [38, 23] but mainly practically that the infection of hundreds of thousands of computers within a matter of hours -or even minutes is feasible. At the theoretical level Staniford [38] presented scanning techniques (random scans, localized scans, hit-list scans, permutation scans) which, used by a worm, can perform attacks of this order. Indeed such worms are often referred to as Warhol worms or Flash worms due to their potential velocity of transmission. A similar confirmation was obtained practically in the cases of the worms Code Red [31], Code Red (CRv2) [5], Code Red II [13], Nimda [26, 27, 20], and Slammer [22], which were characterized as epidemics by the scientific community [44] (although a more appropriate epidemiological term would be pandemics). Recently the Blaster-worm [24, 21] caused significant disruption in the Internet, although the infection rate of the specific worm was relatively slow in comparison with the previously mentioned worms. The reason for the effectiveness of the Blaster-worm was the exploitation of the Windows DCOM RPC interface buffer overrun vulnerability. This vulnerability affects all unpatched Windows NT /2000/ XP systems, as opposed to Code Red worms variations or the Slammer worm which were focused on machines acting as Web Servers or SQL Servers respectively. All of the above is evidence that rapid malcode is extremely hard to confront using the “traditional” way of isolating and studying the code to extract the appropriate signature and update the IDS in real time. We now propose to the reader to consider human behavior during a flu epidemic. Obviously a visit to a doctor


31

and the use of vaccines is essential, however there is also need for an increased awareness and use of hygiene rules: avoiding crowded spaces, increasing the ventilation of our working area etc. Once the epidemic subsides, these measures can be suspended; a person showing symptoms of the disease, of course, should still visit a doctor to receive medical care, regardless of whether the epidemic is still taking place. The classic computer protection methods can be likened to the above medical situation: The vaccination of the population can be compared to updating the virus signature files; the lookout for symptoms may be compared to detection by an IDS; while the hygiene rules followed, which are essential for the protection of the larger, still unaffected population, may be compared to the operation of our proposed system, described in the following sections. 2. Architecture: Peer-to-peer networks, which we will hereafter reference as p2p networks, are often considered to be security threats for organizations, companies or plain users, mainly due to the use of p2p-based applications for illegal file sharing, and to the ability of worms to be spread through such applications (e.g. VBS.GWV.A [41, 40] and W32.Gnuman [10]). Our work indicates, however, that p2p networks can also be positively utilized to significantly reinforce network security, by offering substantial help in the protection against malicious applications. We propose an effective way to achieve this by collecting and exchanging information that will allow us to obtain a global overview of the network status, with reference to ongoing security attacks. The goal of our methodology is to select the most appropriate security policy, based on the level of danger posed by rapid malcode circulating in the network.

P2p networks leverage the principle that a much better utilization of resources (processing power, bandwidth, storage etc.) is achieved if the client/server model is replaced by a network of equivalent peers. Every node in such a p2p network is able to both request and offer services to other peer nodes, thus acting as a server and a client at the same time (hence the term “servent” = SERVer + cliENT which is sometimes used). The motivation behind basing applications on p2p architectures or infrastructures derives to a large extent from their adaptability to variable operating environments, i.e. their ability to function, scale and self-organize in the presence of a highly transient population of nodes (or computers/users), hardware failures and network outages,without the need for a central administrative server.

Our proposed application, which we call “NetBiotic”, requires the cooperation of several computers within a common peer group, in which messages are exchanged describing the attacks received by each computer. It consists of two independent entities: a Notifier and a Handler. These entities act as independent daemons for UNIX systems, services for Windows NT/2000/XP or processes for Windows 9x/Me. From now on we will be referring to these entities as daemons for simplicity. Figure 1 illustrates the architecture of the proposed system within a group of cooperating peer computers.

The Notifier is a daemon responsible for monitoring the computer on which it runs and collecting any information relevant to probable security attacks. There is a plethora of different approaches to incorporate in the Notifier; for simplicity in our preliminary implementation we only monitor the log files of several security related applications, such as firewalls, anti-virus software and IDS systems. These are applications that collect information about security threats and attacks to the computer system on which they are running and either notify the user of these attacks or take specific measures, while at the same time storing information relevant to the attacks into log files. By regularly reading the log files generated by these applications, the Notifier detects any recently identified security attacks to the computer it is running on. At regular time intervals t, the Notifier of node n will record the number of hits (hn

t) the node received over the past interval. It will then calculate and transmit the percentage pn

t by which this average differs from the average hits in an aggregate of the k latest intervals, given by

Figure 1. The architecture of the NetBiotic system

within a group of cooperating peer computers.


32

where: • t is the ordinal number of a fixed time interval. • n is a node identifier. • hn

t is the number of attacks node n received in the interval t.

• pnt is the percentage increase or decrease in attacks

during the current interval t on node n. • k(>0) is the size of the “window” used, in number

of t time intervals, within which the average attack rate is calculated.

Selecting the appropriate length of the time interval t is currently a subject of further research. In our current implementation we use a value of 15 minutes, which we feel provides a balance between increased network traffic and delay in notifying the network of attacks. This will be further discussed in the next Section.

A value of pnt significantly greater than 1.0 is an

indication that node n is under security attack during the interval t. The actual threshold used for pn

t is set by experience, and can vary according to the tolerance for false positives/negatives one has. With a small threshold it is possible to falsely interpret slightly increased rapid malcode activity as an epidemic (false positive), leading to an unnecessary activation of the available countermeasures, which in turn can disrupt some non critical useful services and cause inconvenience to the users. A very large threshold on the other hand, would probably fail to identify a rapid malcode epidemic (false negative) leaving the system protected only by its standard built-in security mechanisms. We tend to believe that is much better to tune the NetBiotic system towards a large threshold because rapid malcode epidemics cause a number of side-effects which are difficult to remain unnoticed. For us it is more important to ensure the timely recognition of these symptoms, in order to increase the security level of the protected system before a circulating worm may manage to launch an attack against it.

The Handler is also a daemon, responsible for receiving the messages sent from the Notifiers of other computers, and for taking the appropriate measures when it is deemed necessary. More specifically, it records the hit rates ht and percentage changes pt received from the different nodes in the peer group within a predefined period of time t, and calculates the overall change in attack rate, averaged for all n nodes of the peer group that transmitted a message during the last interval:

The architecture supports countermeasures based

upon predefined thresholds for pavg,, which are again set by experience. If pavg, exceeds an upper threshold, the security level of the computer is raised. If, on the other hand, it drops below a lower threshold for a large period of time, the security level at which the computer functions is reduced. Selecting the appropriate thresholds τhigh and τlow for increasing or decreasing the security levels is crucial. In our

approach, the thresholds are selected empirically and we have:

• if pavg> τ high then increase security policy. • if pavg< τlow then decrese security policy. • if τlow≤ pavg≤ τ high, do nothing.

We base our decision for modifying the security policy

on the rate of change of attacks, rather than on the actual number of attacks, to normalize the inputs from all peers with respect to their regular susceptibility to attacks; a peer whose actual number of attacks during a monitored time interval has increased from 1000 to 1100 has only experienced a 10% change in the attack rate, while a peer whose number of attacks increased from 50 to 150 within the same interval has experienced a 200% change in the attack rate; still, they have both received 100 attacks more than usual. As far as the actual utilization of our architecture for protecting the computer system is concerned, the countermeasures taken will depend on many factors. A simple personal computer will be requiring different protection strategy than the central server of a large company. The type of operating system is also an important factor. The proposed system is not suggested as a replacement for traditional protection software (anti-viruses, IDS, firewalls etc.). The aim of NetBiotic is to assemble an additional, overall picture of the network status and suggest the basic security measures to be taken in the event of an epidemic. The NetBiotic architecture might not be capable to protect against a specific attack, however it will engage the standard measures that in many cases are crucial (such as disabling HTML previewing in several mail clients, not allowing Active X controls in various web browsers, disabling macros in some office application etc.).

In our prototype design, the recommended measures for a simple personal computer running Microsoft Windows would be to increase the security level of the default mail client and web browser. It would be additionally helpful to alert the user of the increased threat, in order to minimize threats of automated social engineering attacks. Servers can similarly disable non-critical networked services (e.g. by modifying the inetd.conf file in the case of Linux/Unix based operating systems). Figure 2 illustrates the operation and interaction of the Notifier and Handler daemons. 3. Implementation The prototype system we present here was developed using the JXTA protocol [15]. JXTA is a partially centralized p2p protocol implementation introduced in early 2001, designed for maximum peer autonomy and independence. It allows applications to be developed in any language, it is independent of operating system type and is not limited to the TCP/IP protocol for data transfer. This allows an application such as NetBiotic to be easily ported to various operating systems, which is crucial to its operation, as its effectiveness will depend on the size of the peer group that will adopt it. An additional benefit of JXTA is its availability under an open source software license agreement, similar to the Apache License [1]. Due to the nature of our application, security issues are


33

of particular interest. Security provisions are usually incorporated in p2p architectures by means of various cryptographic mechanisms such as the information dispersal algorithm [30] or Shamir’s secret sharing code [33], anonymous cryptographic relays [32], distributed steganographic file systems [11], erasure coding [19], SmartCards or various secure routing primitives [7]. JXTA peers function under a role-based trust model, whereby individual peers function under the authority of third-party peers to carry out specific tasks. Public key encryption of the messages exchanged, which may be in XML format, as well as the use of signed certificates are supported, providing confidentiality to the system. The use of message digests provides data integrity, while the use of credentials — special to-kens that authenticate a peer’s permission to send a message to a specific endpoint —provide authentication and authorization. JXTA also supports the use of secure pipes based on the TLS protocol. Further work is being carried out based on the security issues of the JXTA system, notably the implementation of a p2p based web of trust in the Poblano Project [4], which will be discussed in the future work Section.

Our system was implemented in Java (java2 version 1.4.0 02) using JXTA version 1.0, and uses the winreg [36] tool to administer the windows registry and modify the security settings of the various applications. The main advantages of Java are its compatibility with most operating systems as well as the fact that it is one of the most secure programming languages.

In our preliminary implementation, the Handler modifies the security settings of the Microsoft Outlook mail client and the Microsoft Internet Explorer web browser.These two applications were selected as they are often the target of viruses. The simple operation of increasing their security settings is therefore enough to provide effective protection to a large number of users.

Most anti-virus programs can be adjusted to produce log files with the attacks they intercept. By regularly monitoring such log files, the Notifier daemon is able to detect a security attack and notify the peers. To test our prototype system, we created a software tool which randomly appends supposed security attack entries to these log files. The NetBiotic architecture is compatible with any IDS or anti-virus software that can be setup to record the security attacks against the system it is protecting in a log file. Our aim is to make the NetBiotic system as independent as possible from the IDS with which it cooperates and the underlying operating system. This independence, however, cannot be total, as the following factors will be unavoidably system dependent: ¯ Log files In its simplest form, the system can simply check the size of the log file. For a more sophisticated operation, though, it would be necessary to incorporate a parser that would extract specific information from the log files. Such a parser has to be specific to each different type of log file used. ¯ Countermeasures taken

System independence cannot be achieved in the case of the countermeasures taken, which will depend on the operating system. Different scripts have to be used to modify the security levels of applications in different operating systems.

Figure 2. Operation of the Notifier and Handler daemons Our system has been tested in laboratory environment as well as in a peer group that was set up for this purpose, in which virus attacks were simulated on some peers, resulting in the modification of the security settings of Microsoft Outlook and Internet Explorer on other peer computers. No real viruses were deployed. A program was running on each of the peer computers and periodically edited the log file of the antivirus software, simply changing its size to simulate a security attack event. The average frequency with which these events were simulated was random and different for each computer. The exchange of messages, individual and overall average hit rates as well as the resulting changes in the security settings of the application were recorded and verified against our theoretical expectations. Finally, since our system consists of two independent daemons, it is possible to only install one of the two on certain peer computers. For instance, the Notifier daemon would be particularly useful running on a large company server, and supplying the peers with information about the security threats it faces. The administrators of such a server may prefer not to install the Handler daemon, and instead manually take action in the event of security attacks. Similarly, for a personal computer user who may not have adequate security measures and antivirus software installed (for either financial or other reasons), installing the Handler


34

daemon itself may provide an adequate level of protection. In this case, the Handler daemon would modify the local security level based on information received by the security focused peer group. The Handler would therefore operate relying on the trustworthiness of the information received from the peer group only, which may in some cases be a disadvantage.

4. Related Work

The research that is most relevant to our proposed system has been carried out within the framework of project Indra [14], with which we partially share a common philosophy. We agree on the basic principle of using p2p technology to share security attack information between computers in a network in order to activate security countermeasures if necessary. We differ however in the circumstances under which specific countermeasures should be taken. According to the Indra project team, in the event that a security attack is detected countermeasures should be immediately initiated, by using the appropriate plugins to protect the computer system. A single security attack anywhere in the network is enough for them to generate a response. In short, Indra is designed to respond to every single security attack. In contrast, our system’s goal is to determine if there is a general increase in the virus or worm attacks in the network, or more importantly a virus or worm epidemic outbreak. Measures taken in this case, such as the increase in security settings of mail clients, web browsers and anti-virus programs will only be effective during the epidemic, and the system will return to its original state after it is finished. In our design, individual virus or worm attacks in the network are not considered separately. Furthermore, we believe that our design can be expanded to very large network sizes without considerably increasing the overall network traffic. A number of highly distributed systems rely on peer communications. The Hummingbird system [28] is based on a cooperative intrusion detection framework that relies on the exchange of security related information between networks or systems in the absence of central administration. The structure of the Hummingbird system is significantly more complex and advanced than NetBiotic, using a combination of Manager-Hosts, Managed Hosts, Slave Hosts as well as Peer, Friend and Symbiote relationships for the exchange of security related information. The Hummingbird system includes advanced visualization tools for its configuration and monitoring of log files, and although it may require considerable effort and expert knowledge for fine tuning the cooperation of each host with the others, it is particularly effective for distributed security attacks (such as doorknob, chaining, loopback attacks etc.). A potential secondary use of the Hummingbird system, in our view, could also be in the detection of malcode.

Emerald [29, 25] is a system targeted towards the exchange of security incident related information between different domains or large networks. It consists of a layered architecture that provides a certain abstraction, and requires the adjustment of parameters relevant to the trust relationships between cooperating parties. We believe that

Emerald, like Hummingbird, can be invaluable in protecting a computer system or network against distributed and targeted attacks. NetBiotic may not be in the position to affront such attacks with the same effectiveness, as its goal is the seamless and automated creation of a network of peers for the fast exchange of information regarding rapid spread malcode activity, leveraging the benefits of peer-to-peer architectures and topologies, and providing basic protection to the participating peers.

Bakos and Bert [2] presented a system for the detection of virus outbreaks. The fastest spreading worms use scanning techniques for identifying potential target computers. As a result, they also scan a large number of addresses that do not correspond to actual computers. The routers that intercept such scanning messages usually reply with a ICMP Destination Unreachable (also known as ICMP Type 3 or ICMP-T3) message. The authors propose that a carbon copy message be sent by the routers to a central collector system, which will be responsible for collecting, correlating and analyzing this data. Bakos and Bert have implemented such a system by modifying the kernel of the Linux operating system to act as a router. The central collector receives the messages and forwards them to an analyzer system, which extracts the valuable information. It should however be examined whether the time required for the entire processing prohibits the use of this system for fast spreading worms, as described by Staniford [38].

Systems that use an extended network to gather information yet rely on a centralized client/server model were also examined. DeepSight [6] is a system developed by Symantec based on a client/server architecture, whereby centralized servers collect and re-distribute security attack information. Since it is a commercial system it is not available for scientific research, however it does include a very widespread data collection network. An approach similar to DeepSight is taken by DShield, in which hundreds of computers communicate with central servers and transmit their IDS log files. The servers process the data and announce in a web site information about the currently active malware, the IP addresses from which most attacks originated and other useful information. Through the incorporation of different parsers, DShield supports various different IDS systems. DShield has been active for more than two years, with a significant number of users. A disadvantage of the system is that the large volume of data collected requires considerable processing time for extracting useful information. The theoretical times taken by the Flash and Warhol worms as well as the measured times for the Slammer worm [22, 38] to spread through the Internet are probably beyond the ability of DShield to react. Both DeepSight and DShield aim at providing a global view of the Internet security status, however they are both subject to the disadvantages of the client/server architecture they follow: their dependence on a single server for their operation and their lack of adaptability makes them vulnerable to targeted attacks. An original approach taken by the AAFID [35], whereby agents are used to collect virus attack information also follows a


35

centralized control structure. The same holds for the GrIDS system [39], which uses activity graphs to control large scale networks and identify suspicious activities, based on the judgment of a System Security Officer. Finally, the following two approaches propose different ways of monitoring the overall security state and threat level of a network: In the DIDS system [34], the overall security state of a network under observation is represented by a numerical value ranging between 0 (safest) and 100 (least safe), while a clearly visual approach to representing the network security state has been proposed [42, 8]. We find both approaches very descriptive and useful to a System Security Officer. In our prototype NetBiotic implementation, however, we are currently adopting a much simpler approach which consists of choosing between three different security states (regular, low risk and high risk), as described in Section 2. 5. Future Work

The NetBiotic system is an evolving research prototype. It is currently being extended in a number of ways as discussed below, in order to subsequently be released as open source software to allow the collaboration with other research groups working in similar directions. At this stage, our goal is to propose an architecture, accompanied by a basic implementation for proof-of-concept purposes, which, based on a p2p network infrastructure can provide security services for computer systems. Although our prototype performed well in the situation in which we tested it, it is not suitable for performing large-scale testing. We expect that, before more advanced versions of our application will be implemented, the scientific community will examine the use p2p networks in security applications from a theoretical standpoint and provide insight into the advantages and disadvantages of such an approach. The following conceptual and implementation improvements are currently being considered: ¯ Vulnerability to malicious attacks A major drawback of our current design is its inability to effectively verify theinformation transmitted in the network. If one or more malicious users manage to introduce in the peer network a large number of false hit rate indications, the result may be the unwanted decrease of the security measures of the computers in the network, rending them vulnerable to virus attacks. We propose that all members of the security peer group will have to be authenticated and verified, probably through the use of certificates, to enforce a consistent authentication and authorization policy. At the implementation level, to confront the problem of malicious users introducing false information we further propose the following approaches, based on the capabilities offered by JXTA: 1. JXTA supports the exchange of encrypted messages based on the TLS algorithm secured pipes [3], which will be used for the transmission of warning messages. 2. JXTA message digest will be used for data integrity

purposes. 3. Other research groups are involved in the creation of a p2p-based web of trust. We intend to study these systems to examine to what extent they can be used to enhance the NetBiotic architecture. ¯ Use of epidemiological models We believe that the incorporation of mathematical epidemiological models for the detection of epidemic outbreaks in the network and determining the threshold for initiating security level modifications should significantly enhance the robustness of our system. A key point in our future research will be the selection of the thresholds for modifying security policies. These thresholds will be variable and will depend on each system’s characteristics and on an analysis of the attack data collected. Studies [9, 18, 16, 17] show that there is a correlation between the patterns of spread of biological viruses and computer viruses. These studies were mainly limited to closed local area networks. P2p models are ideal for gathering large scale network virus information, which can subsequently be processed and adapted to epidemiological models, leading to decision tools for concluding, or perhaps even predicting, whether there is — or is likely to be — an epidemic outbreak in the network. ¯ Choice of appropriate security policy In conjunction with other factors, such as the role of the system being protected, our system should be able to effectively choose the most appropriate security policy for the specific period of time. In this way, single incidents of virus attacks may not be the cause of any concern, yet the detection of epidemic outbreaks would initiate a modification of the security policies. ¯ Platform porting In porting our system to Unix/Linux platforms, the operating system could be instructed to launch or halt applications, or automatically request updates. The configuration of these operating systems can be edited through plain text files, which is an additional benefit for our system. 6. Conclusions

Even the best protected organizations, companies or personal users are finding it difficult to effectively shield themselves against all malicious security attacks due the increasing rate with which they appear and spread. Antivirus applications, as well as IDS systems, identify the unknown malware by employing behavioral based heuristic algorithms. These algorithms are particularly effective under a strict security policy, however they tend to produce an increased number of false alarms, often disrupting and upsetting the smooth operation of a computer system and the organization or users it supports. On the other hand, if the security policy is relaxed, the threat of a virus infection becomes imminent. We propose a platform based on p2p technology in which the computers participating as peers of a network automatically notify each other of security threats they receive. Based on the rate of the warning messages received, our system will increase or decrease the security measures taken by the vulnerable applications running on


36

the computer. Our approach automates elements of the process of choosing the appropriate security policy, based on data useful for adjusting the security levels of both the operating system (by launching and terminating related applications) and the security applications (by modifying the security parameters of the heuristic algorithms they employ). An important aspect of our design is that the traffic introduced in the network by the peer nodes as a result of the transmission of hit rate information is minimal. We believe that, with the inclusion of the future extensions we are currently working on, our approach may lead to operating systems, antivirus programs, IDS software and applications that will be able to self-adjust their security policies. References: [1] Apache license: Current on-line (June 2003):

http://httpd.apache.org/docs/license. [2] G. Bakos and V. Berk. Early detection of internet

worm activity by metering icmp destination unreachable messages. In Proceedings of the the SPIE Aerosense, 2002.

[3] Wilson B.J. JXTA. New Riders, Indianapolis, IN, USA, June 2002.

[4] R. Chen and W. Yeager. Poblano: A distributed trust model for peer-to-peer networks. Technical report, Sun Microsystems.

[5] Code Red CRv2. Current on-line (June 2003): http://www.caida.org/analysis/security/code-red/coderedv2 analysis.xml.

[6] Deepsight threat management system: Current on-line (June 2003): http://www.securityfocus.org.

[7] P. Druschel and A. Rowstron. Past: A large-scale, persistent peer-to-peer storage utility. In Proceedings of the Eighth Workshop on Hot Topics in Operating Systems, May 2001.

[8] R. Erbacher, K. Walker, and D. Frincke. Intrusion and misuse detection in large scale systems. IEEE Computer Graphics and Applications, 22(1), 2002.

[9] S. Forrest, S. Hofmeyr, and A. Somayaji. Computer immunology.Communications of the ACM, 40(10):88–96, 1997.

[10] W32.gnuman.worm: Current on-line (June 2003): http://service1.symantec.com/sarc/sarc.nsf/html/w32.gnuman.worm.html.

[11] S. Hand and T. Roscoe. Mnemosyne: Peer-to-peer steganographic storage. In Proceedings of the 1st International Workshop on Peer-to-Peer Systems (IPTPS ’02), MIT Faculty Club, Cambridge, MA, USA, March 2002.

[12] Icsa labs 2002 computer virus prevalence survey. Current on-line (June 2003): http://www.trusecure.com/download/dispatch/vps2002.pdf.

[13] Code Red II. Current on-line (June 2003): http://www.eeye.com/html/research/advisories/al20010804.html.

[14] R. Janakiraman, M. Waldvogel, and Q. Zhang. Indra: A peer-to-peer approach to network intrusion detection and prevention. In Proceedgings of 2003

IEEE WET ICE Workshop on Enterprize Security, Linz, Austria, June 2003.

[15] Project jxta v2.0 java programmer’s guide: Current on-line (June 2003): http://www.jxta.org/docs/jxtaprogguide v2.pdf.

[16] J Kephart. How topology affects population dynamics. In Proceedings of Artificial Life 3, Santa Fe, New Mexico, June 1992.

[17] J. Kephart, D. Chess, and S. White. Computers and epidemiology. IEEE Spectrum, May 1993.

[18] J. Kephart and S. White. Directed-graph epidemiological models of computer viruses. In Proceedgings of IEEE Computer Society Symposium on Research in Security and Privacy, pages 343–361, Oakland, CA, 1991.

[19] J. Kubiatowicz, D. Bindel, Y. Chen, P. Eaton, D. Geels, S.R. Gummadi, H. Weatherspoon, W. Weimer, C. Wells, and B. Zhao. Oceanstore: An architecture for global-scale persistent storage. In Proceedings of ACM ASPLOS. ACM, November 2000.

[20] A. Mackie, J. Roculan, R. Russell, and VanVelzen M. Nimda worm analysis - incident analysis report version ii. September 2001.

[21] J. Miller, J. Gough, B. Konstanecki, J. Talbot, and J. Roculan. Deepsight threat management system threat alert - microsoft DCOM RPC worm alert. Current on-line (August 2003):

https://tms.symantec.com/members/analystreports/030811-alert-dcomworm.pdf.

[22] D. Moore, V. Paxson, S. Savage, C. Shannon, S. Staniford, and N. Weaver. The spread of the sapphire/slammer worm. Current on-line (June 2003): http://www.caida.org/outreach/papers/2003/sapphire/sapphire.html. Technical

report, 2003. [23] D. Moore, G. Voelker, and S. Savage. Internet

quarantine:requirements for containing self-propagating code. In Proceedings of the 2003 IEEE Infocom Conference, San Francisco California, USA, April 2003.

[24] Microsoft security bulletin ms03-026. Current on-line (August 2003): http://www.microsoft.com/technet/treeview/default.asp? url=/technet/security/bulletin/ms03-026.asp.

[25] P. Neumann and P. Porras. Experience with EMERALD to date. In First USENIX Workshop on Intrusion Detection and Network Monitoring, pages 73–80, Santa Clara, California, April 1999.

[26] Current on-line (June 2003): http://www.incidents.org/react/nimda.pdf.

[27] Current on-line (June 2003): http://www.f-secure.com/v-descs/nimda.shtml.

[28] Polla, D., J. McConnell, T. Johnson, J. Marconi, D. Tobin, and D. Frincke. A framework for cooperative intrusion detection. In Proceedings of the 21st National Information Systems Security Conference, pages 361–373, October 1998.

[29] P. Porras and P. Neumann. EMERALD: Event monitoring enabling responses to anomalous live


37

disturbances. In Proceedings of the National Information Systems Security Conference, October 1997.

[30] M.O. Rabin. Efficient dispersal of information for security, load balancing and fault tolerance. Journal of the ACM, 36(2):335–348, April 1989.

[31] Code Red. Current on-line (June 2003): http://www.eeye.com/html/research/advisories/al20010717.html.

[32] A. Serjantov. Anonymizing censorship resistant systems. In Proceedings of the 1st International Workshop on Peer-to-Peer Systems (IPTPS ’02), MIT Faculty Club, Cambridge, MA, USA, March 2002.

[33] A. Shamir. How to share a secret. Communications of the ACM, 22:612–613, November 1979.

[34] S. Snapp, J. Brentano, G. Dias, T. Goan, T. Heberlein, C. Ho, K. Levitt, B. Mukherjee, S. Smaha, T. Grance, D. Teal, and D. Mansur. DIDS (distributed intrusion detection system) - motivation, architecture, and an early prototype. In

Proceedings of the 14th National Computer Security Conference, pages 167–176, Washington, DC, 1991.

[35] E. Spafford and D. Zamboni. Intrusion detection using autonomous agents. Computer Networks, (34):547–570, October 2000.

[36] D. Spinellis. Outwit: Unix tool-based programming meets the windows world. In Proceedings of the USENIX 2000 Technical Conference, pages 149–158, San Diego, CA, USA, June 2000.

[37] D. Spinellis. Reliable identification of bounded-length viruses is np-complete. IEEE Transactions on Information Theory, 49(1):280–284, January 2003.

[38] S. Staniford, V. Paxson, and N. Weaver. How to own the internet in your spare time. In Proceedings of the 11th USENIX Security Symposium, 2002.

[39] S. Staniford-Chen, S. Cheung, R. Crawford, M. Dilger, J. Frank, J. Hoagland, K. Levitt, C. Wee, R. Yip, and D. Zerkle. GrIDS – A graph-based intrusion detection system for large networks. In Proceedings of the 19th National Information

Systems Security Conference, 1996. [40] VBS.Gnutella. Current on-line (June 2003):

http://service1.symantec.com/sarc/sarc.nsf/html/vbs.gnutella.html.

[41] VBS.Gnutella. Current on-line (June 2003): http://vil.nai.com/vil/content/v 98666.html.

[42] G. Vert, J. McConnell, and D. Frincke. A visual mathematical model for intrusion detection. In Proceedings of the 21st National Information Systems Security Conference, pages 329–337, October 1998.

[43] C. Wang, J.C. Knight, and M.C. Elder. On computer viral infection and the effect of immunization. In Annual Computer Security Applications Conference (AC-SAC), pages 246–256, December 2000.

[44] V. Yegneswaren, P. Barford, and J. Ullrich. Internet intrusions: Global characteristics and prevalence. In Proceedings of ACM SIGMETRICS, June 2003.

[45] R.L. Ziegler. Linux Firewalls. New Riders Publishing, Indianapolis IN, USA., 2002.

Authors Profile

Dr. G.Srinivasa Rao, M.Tech, Ph.D, Sr.Asst.Professor. four years industrial experience and over 8 Years of teaching experience with GITAM University, handled courses for B.Tech, M.Tech. Research areas include Computer Networks And Data Communications. published 6 papers in various National and International Conferences and Journals. Dr. G.Appa Rao., M.Tech., M.B.A.,Ph.D., in computer science and Engineering form Andhra Universiy. Over 12 Years of teaching experience with GITAM University, handled courses for B.Tech, M.Tech. Research areas include Data Mining and AI. Published 8 papers in various National and International

Conferences and Journals.

Mrs. S. Venkata Lakshmi M.Tech in Information Technology from Andhra University. Asst.Prof in GITAM University. Over 2 years of teaching experience with GITAM University and Andhra University handled courses for B.Tech, and M.C.A. and 2 years of industry experience as a software engineer. Published 2 papers in various

International Conferences and Journals. Mr..D.Veerabhadra Rao., M.Tech., in Information Technology. Over 7 Years of teaching experience with GITAM University, handled courses for B.Tech and M.Tech .One research paper was published in international journal and one conference.

B. Venkateswar Reddy M.Sc (Maths) from Osmania University and M.Tech(CS) from Satyabhama University, having one year teaching experience, handled courses for B.Tech and M.Tech in CVSR Engineering college, Hyderabad.

P. Venkateswara Rao M.Sc(Physics) from Acharya Nagarjuna University, Pursuing M.Tech (CST) from GITAM University K. Sanjeevulu M.Sc(Maths) from Osmania University., Pursuing M.Tech (CST) from GITAM University.


38

New Power Aware Energy Adaptive protocol with Hierarchical Clustering for WSN

Mehdi Golsorkhtabar1, Mehdi Hosinzadeh2, Mir Javad Heydari1, Saeed Rasouli1

1 Islamic Azad University, Tabriz Branch, Department of Computer Engineering,

Tabriz, Iran {m.golsorkhtabar;m.heydari;s.rasouli}@iaut.ac.ir

2 Science and Research Branch, Islamic Azad University,

Tehran, Iran [email protected]

Abstract: In recent years, progress in wireless communication has made possible the development of low cost wireless sensor networks. Clustering sensor nodes is an effective topology control approach. In this paper, we propose a new routing algorithm which can increase the network lifetime. We assume that each node can estimate its residual energy and then a new clustering method will be proposed for increase of network lifetime. This assumption is similar to many other proposed routing algorithms for WSN and is a possible assumption. In the new algorithm, the predefined numbers of nodes which have the maximum residual energy are selected as cluster-heads based on special threshold value first and then the members of each cluster are determined based on the distances between the node and the cluster head and also between the cluster head and base station. At last, the simulation results show that our method achieves longer lifespan and reduce energy consumption in wireless sensor networks.

Keywords: wireless sensor networks; clustering algorithm; energy adaptive; network lifetime

1. Introduction Wireless Sensor Network (WSN) comprises of micro

sensor nodes, which are usually battery operated sensing devices with limited energy resources. In most cases, replacing the batteries is not an option. [1-2].

Wireless sensor network, usually, are heterogeneous. The protocols should be design for the typical of heterogeneous wireless sensor networks. Most of the clustering algorithms are designed for homogeneous wireless sensor networks and they are not optimized when network's nodes is in heterogeneous state such as [3-4].

In this paper, we propose and evaluate PEAP, (new Power aware Energy adaptive Protocol with hierarchical clustering for WSN). In considered wireless sensor network, nodes send sensing information to a cluster-head and then the CH transmit data to base station. The certain clustering algorithms with special method periodically electing cluster-heads then cluster-heads aggregate the data of their cluster nodes and send it to the base station. We assume that all the nodes of the network are spread heterogeneous, at first all nodes battery power is equal, all sensor nodes have limited energy and the base station is fixed and not located between sensor nodes and most of them are static and only a few are mobile.

2. Related work In recent years, many protocol and algorithm for clustering have been proposed. Usually there are many differences between CH selection and clustering organism. But generally some of these clustering schemes applied in homogeneous networks and some others clustering algorithms applied in heterogeneous networks. Therefore, most of the current popular clustering algorithms are not fault tolerant, such as LEACH [3], PEGASIS [4] and HEED [5]. LEACH is the most popular clustering algorithm. Many of CH selection algorithms are based on LEACH’s architecture. [6] is proposed to elect the CHs according to the energy remaining in each node. We call this clustering protocol LEACH-E. In the rest of this section, we review LEACH algorithm and discuss its limitations, because LEACH is very popular in wireless sensor network clustering protocols.

2.1 LEACH (Low Energy Adaptive Clustering Hierarchy)

In LEACH protocol, energy efficiency are achieved by being CH in turn, and then distributing impartially the total networks energy to unique node, thus lowing energy consuming and increasing network lifespan. CH election depends on the whole numbers of CH in networks and times that nodes have been CH until now. Principles scenarios for this protocol are:

• The base station fixed in nowhere near of the sensor nodes.

• All the nodes in the wireless sensor network have the same initial battery power and are homogeneous in all other ways.

In first phase, algorithm chooses a node stochastically, the principal will be explained in the following: all sensor nodes compute a value T(n ) according to the following formula at the beginning of all rounds:

) 1(

∈

−=

others

Gnprp

pnT

0

))1mod((1)(

Where in this equation P describes desired percentage of


39

CHs (e.g. P=0.05) current round, and G is the set of nodes that have not been CH in the last 1/P rounds, r is the number of the current round.

For each node, a random number between 0 and 1 is generated. If this random number is less than T(n ) , this sensor node will become a cluster head in this round and broadcast an advertisement message to other sensor nodes near it.

When each node has elected as cluster head itself for the current round broadcasts an advertisement message to the rest of the nodes in the network. All the non-cluster head nodes, after receiving this advertisement message, decide on the cluster to which they will belong for this round. This decision is based on the received signal strength of the advertisement messages. After cluster head receives all the messages from the nodes that would like to be included in the cluster and based on the number of nodes in the cluster, the cluster head creates a TDMA schedule and assigns each node a time slot when it can transmit [3].

Despite many advantages in using of the LEACH protocol for cluster organization, CH selection and incising network lifetime, there are a few features that the protocol does not support. LEACH assumes nodes power energies homogeneously. In a real, wireless sensor networks scenario, sensor nodes energy spread in heterogeneous manner.

3. The New Protocol In this section, the details of PEAP are introduced. The major application of a wireless sensor network is to monitoring of a remote environment. Data of individual nodes are usually not very important. Since the data of sensor nodes are correlated with their neighbor nodes, data aggregation can increase reliability of the measured parameter and decrease the amount of traffic to the base station. PEAP uses this observation to increase the efficiency of the network. In order to develop the PEAP, some assumptions are made about sensor nodes and the underlying network model. For sensor nodes, it is assumed that all nodes are able to transmit with enough power to reach the BS if needed, that the nodes can adjust the amount of transmit power, and each node can support different Medium Access Control (MAC) protocols and perform signal processing functions. These assumptions are reasonable due to the technological advances in radio hardware and low-power computing [3]. For the network, it is assumed that nodes have always data to send to the end user and the nodes located close to each other have correlated data.

Such as LEACH, in first phase, PEAP chooses a node stochastically, , the principal will be explained in the following: all sensor nodes compute a value T(n ) according to the following formula at the beginning of all rounds.

)2(

−=max_

_1*max_

_*)(nE

currentnE

nEcurrentnE

pnT

Where in this equation P = the desired percentage of CHs (e.g. P=0.05) the current round, and En_current is the current energy and En_max the initial energy of the node, with rs as the number of consecutive rounds in which a node has not been CH. Thus, the chance of node n to become cluster head increases because of a higher threshold. A possible blockade of the network is solved. Additionally, rs is reset to 0 when a node becomes CH. Thus, we ensure that data is transmitted to the base station as long as nodes are alive [6].

Our clustering model is based on confidence value associated with broadcast from CHs. Confidence value of a CH is a function of some parameters (1) distance between the CH, the node and (2) the CH current battery power and (3) number of nodes already were a member of this CH. Basically, our model checks first if, with the current battery power the CH has, it would be able to support the current members at maximum data broadcast rate. A node decides to join a CH if the head can still support the node with its rest power. Confidence value given by:

Where in this equation BP is the battery power of given node, Cm is number of nodes already a member of given CH, Dc is distance between the CH and the node.

Like LEACH, in order to reduce the probability of collision among joint-REQ messages during the setup phase, CSMA (Carrier Sense Multiple Access) is utilized as the MAC layer protocol. When a cluster head has data to send, it must sense the channel to see if anyone else is transmitting using the BS spreading code. If so, the cluster head waits to transmit the data. Otherwise, the cluster head sends the data using the BS spreading code [3].

4. Simulation Results In order to evaluate the performance of the PEAP protocol, the simulator, specific to the needs of our model, was coded in PHP with Apache HTTP server version 2.2 and uses PHP/SWF Charts for its graphical needs.

We assume a simple model for the radio hardware energy dissipation where the transmitter dissipates energy to run the radio electronics and the power amplifier, and the receiver dissipates energy to run the radio electronics, as shown in Fig. 1. For the experiments described here, both the free space (d2 power loss) and the multi path fading (d4 power loss) channel models were used, depending on the distance between the transmitter and receiver [7]. Power control can be used to invert this loss by appropriately setting the power amplifier. If the distance is less than a threshold do, the free space (fs) model is used; otherwise, the multi path (mp) model is used. Thus, to transmit l-bit message a distance, the radio expends

)3(

≥+

<+=+=

o4

mpelec

o2

fselecamp-Txelec-TxTx dd ,dllE

dd ,dllEdlElE d)(l,E

ε

ε),()(

And to receive this message, the radio expends:

Dc* CmBp

v (i)C = (3)


40

)4( elecelec-RxRx EllE (l)E == )(

The electronics energy, Eelec, depends on factors such as the digital coding, modulation, filtering, and spreading of the signal, whereas the amplifier energy, εfsd2 or εmpd4, depends on the distance to the receiver and the acceptable bit-error rate.

We consider a wireless sensor network with N=100 nodes randomly distributed in a 300m * 300m field. We assumed the base station to be fixed and located at the origin (0, 0) of the coordinate system. The radio parameters used in our simulations are shown in Table (1).

Figure 1.Total nodes transmitting (N= 100).

Table 1: Parameters used in simulations Parameter Value

Eelec 50 nJ/bit

ε£s 10 pJ/ bit/m2

εmp 0.0013 pJ/bit/m4

E0 3 J

EDA 5 nJ/bit/message

d0 70 m

Message size 8192 bits

Fig. 2 present the energy consumption of the clustering protocols when the amount of nodes spread in network is 100. The x-axis indicates the number of rounds while y-axis indicates the mean residual energy of each node. The results demonstrate that the energy consumption of our algorithm is generally smaller than LEACH and LEACH-E.

Figure 2.Total network energy (N=100)

Fig. 3 presents the number of nodes dead when using of clustering protocols. This result is closely related to the network lifetime of the wireless sensor networks.

Figure 3.Total nodes transmitting (N=100).

5. Conclusion In wireless sensor networks, the energy consumption and the network lifetime are important issues for the research of the route protocol. This paper introduces PEAP, a new power aware energy adaptive protocol with hierarchical clustering for wireless sensor networks that distributes loads among more powerful nodes. Compared to the existing clustering protocols, PEAP has better performance in CH election and forms adaptive power efficient and adaptive clustering hierarchy. The simulation results presented that PEAP significantly improves the lifespan and the energy consumption of the wireless sensor networks compared with existing clustering protocols. Further directions of this study will be deal with clustered sensor networks with more than three parameters with in threshold calculating and more parameters to confidence value calculating.

References [1] C. Buratti, A. Conti, D. Dardari,R. Verdone, “An

Overview on Wireless Sensor Networks Technology and Evolution” , Sensors 2009, 9, 6869-6896; doi:10.3390/s90906869.

[2] I. F. Akyildiz, W. Su, and Y. Sankarasubramaniam, “A survey on sensor networks”, IEEE Communications Magazine, 2002, 40(8), pp.102-114.

[3] W.R. Heinzelman, A.P. Chandrakasan, H. Balakrishnan, “An application-specific protocol architecture for wireless microsensor networks”, IEEE Transactions on Wireless Communications 1 (4) (2002) 660–670.

[4] S. Lindsey, C.S. Raghavenda, “PEGASIS: power efficient gathering in sensor information systems”, Proceeding of the IEEE Aerospace Conference, Big Sky, Montana, March 2002.

[5] O. Younis, S. Fahmy, “HEED: A hybrid, energy-efficient, distributed clustering approach for ad hoc sensor networks”, IEEE Transactions on Mobile Computing 3 (4) (2004) 660–669.

[6] M.J. Handy, M. Haase, D. Timmermann,“Low energy clustering hierarchy with deterministic cluster head selection”, Proceedings of IEEE MWCN, 2002.

[7] T. Rappaport, Wireless Communications: Principles & Practice. Englewood Cliffs, NJ: Prentice-Hall, 1996.


41

Hexagonal Coverage by Mobile Sensor Nodes

G.N. Purohit1, Seema Verma2 and Megha Sharma3

1Department of Mathematics, AIM & ACT, Banasthali University, Banasthali-304022

[email protected]

2Department of Electronics, AIM & ACT, Banasthali University, Banasthali- 304022

[email protected]

3Department of Computer Science, AIM & ACT, Banasthali University, Banasthali-304022

[email protected] Abstract: Before the advent of mobile sensor nodes, static nodes have been used to provide coverage, which focuses on repositioning of sensors to achieve coverage. But mobile sensor nodes provide a dynamic approach to cover age. Targets that might never be detected in a stationery sensor network can be detected by moving sensors. Mobile sensors can compensate for lack of sensors and improve network coverage. Here, we focus on coverage of a rectangular region which is divided into regular hexagons. The region is covered with mobile sensor nodes, where a group of four MSNs position themselves on four vertices of a hexagon. We can employ N≥4 MSNs for this purpose, although basically only 4 MSNs are needed but extras are employed in case of failure of any MSN. Key Words: coverage, mobile sensor nodes, energy efficiency, hexagonal coverage.

1. Introduction The coverage problem is a fundamental issue in WSN, which mainly concerns with a fundamental question: How well a sensor field is observed by the deployed sensors? To optimize network coverage, the traditional approach is to deploy a large number of stationary sensor nodes and then to schedule their sensing activities in an efficient way [6]. Recently, mobile sensor nodes have received much attention since network performance can be greatly improved by using just a few of mobile nodes. Mobile sensor nodes have the movement capability to collaboratively reinstall the network coverage. They are extremely valuable in situations where traditional deployment mechanisms fail or are not suitable, for example, a hostile environment where sensors cannot be manually deployed or air-dropped. It is well known that mobility increases the capacity of networks (MANETs) by reducing the number of relays for routing, prolonging the lifespan of wireless sensor networks (WSNs) and ensuring network connectivity in delay-tolerant networks (DTNs), using mobile nodes to connect different parts of a disconnected network. In this paper we present Mobile Traversal Algorithm (MTA) where the region of interest [ROI], considered as a rectangular area, is covered by regular hexagons. The MSNs are placed at four vertices

of the hexagon. They move in a systematic manner on rectangular and triangular parts of hexagon. Previous work on Mobile traversal has been done using triangulation based coverage [3], but the hexagonal approach proves to be more efficient as the total distance traveled and time taken is comparatively less. Deploying a good topology is also beneficial to management and energy saving, and the hexagonal topology provides 2-coverage, as we wish to ensure optimal and energy efficient coverage. A deterministic energy-efficient protocol for sensor networks is used in [1] that focuses on energy efficient coverage of ROI. Energy efficient distributed algorithms for sensor target coverage based on properties of an optimal schedule is included in [2]. Power efficient organization of wireless sensor networks is done in [4]. A coverage-preserving node scheduling scheme for large wireless sensor networks is discussed in [5]. The proposed objectives of our approach are:

(i) Covering the sensing area by minimum number of sensors, N≥4, as well as providing highly reliable and long system lifetime, which is the main design challenge in sensor networks.

(ii) Upon a failure, the remaining MSN’s (N-4) efficiently complete the coverage of the targeted area, otherwise they remain in sleeping mode.

We assume the following: (i) The sensing range of a sensor x, is a disc of radius r centered at x and defined by Sx (r) = {a € R2 : | x-a | ≤ r} where |x-a | stands for the Euclidian distance between x and a.

(ii)A location in region A is said to be covered by sensor x if it is within x’s sensing range. A location in A is said to be covered if it is within at least K sensor’s sensing range.


42

(iii) All sensors are considered to have identical sensing range. This paper is organized as follows. The problem is formulated in Section 2. In Section 3 the MTA is presented, and the total traveling distance for various sides ‘a’ of MSNs is determined. In Section 4 we conclude the paper.

2. Problem Formulation We consider a rectangular field of length ‘L’ and breadth ‘B’ which is divided into regular hexagons of sides a, We consider all sensors having uniform sensing range r and side of the regular hexagon a is taken less than r. The hexagon is further subdivided into three parts: two isosceles triangles and a rectangle, as shown in Fig1. The whole rectangular area is covered by m*n hexagons.

Figure 1. Division of rectangular field into regular

hexagons

However, the area is over covered by some hexagons covering the perimeter of the ROI, shown as the shaded area in Figure 1. The length (L) and breadth (B) of the targeted region are related with m, n and side a of the hexagon by following relations: L = a (m-1) √3, B = a (2n-1) Thus there are n rows of hexagons and in each row there are m hexagons.

The rows are numbered 1 to n and the columns are numbered 1 to m. For sake of convenience we consider the center of the top most left hexagon as origin of reference, x-axis along horizontal line and y-axis as the vertical line downwards The coordinate positions of the centers of hexagons dividing the rectangular field are represented as (x, y) = (a√3i, aj), which represent the position of the centre of the hexagon in ith row and jth column, where x= 1…n, y = 1…m and 1≤i≤n, 1≤j≤m.

We can see that each point in the ROI is covered by at least 2 sensors, which is depicted by the figure shown below

Figure 2. K-Coverage of a point

3. MTA (Mobile Traversal Algorithm) The side of regular hexagons is taken less than or equal to the sensing range of the sensors. i.e., a ≤ r. Given N ≥ 4 MSN’s, we take four sensors, which are placed at a (0, 0), b (0, a), c (a√3/2, a), d (0, a√3/2), and the sensors move as indicated in Figure 3. Considering only the hexagons in the ROI i.e., the rectangular region, the 4 MSNs a, b, c, d move towards the right until the last column is reached.

Figure 3. Traversal schemes of MSNs

Out of these four MSNs only three MSNs cover the triangles below the rectangles of the hexagon in first row. After reaching to the leftmost triangle of the first row these three MSNs cover the top triangles of the hexagons in the second row until they cover up to the right most top triangle of second row. Then these three MSNs along with the remaining MSNs position on the four vertices of the right most half rectangle of the second row and cover all the rectangles as done earlier in first row. This process continues until the whole unshaded area in the Figure 3 is covered. The distance traveled by these MSNs covering rectangles, and triangles, rectangles in the indicated directions are detailed in the next paragraph. The time taken is proportional to distance covered by the MSNs.

In the first movement only two MSNs ‘a’ and ‘b’ together travel a distance of 3√3a to cover the first full rectangle, Fig 4(a). To cover the remaining full rectangles the 4 MSNs together move a total distance of 4√3a in each move, to cover a rectangle, Fig 4(b). To cover the last right most top half-rectangle the MSNs ‘a’ and ‘b’ together travel a distance of 3√3a, Fig 4(c). To cover the right-angled (right-most half triangle in the first row) triangle, two MSNs (namely ‘d’ and ‘c’ in figure) out of these 4 MSNs are kept stationery and the other two move along the


43

indicated directions and travel together a distance 2a, Fig 4(d). To cover the isosceles triangle one MSN (namely ‘c’ in figure) moves along the indicated direction a distance of √13/2a, Fig 4(e). To cover the next isosceles triangles MSN ‘b’ travels a distance of √7a, Fig 4(f). To cover the right-angled triangle(right most half triangle) in the next row MSNs ‘a’, ‘b’, ‘c’ travel a distance of 3a, Fig 4 (vii) To cover the rightmost half-rectangle in the next row the four MSNs (namely ‘a’, ‘b’, ‘c’, ‘d’ in figure) travel a total distance of 5a , Fig 4 (g).This way the MSNs cover the rectangular region (ROI) upto m columns and n rows.

Figure 4. MSN movements between hexagons

3.1 MTA with failure tolerance In order to provide failure tolerance to MTA described above, we add a few extra MSN’s which in the failure of a particular MSN would occupy its position, otherwise these extra MSN’s stay in the sleep mode.

Figure 5. Extra sensors to provide coverage in case of

sensor failure

If a sensor fails at say ith row and jth column, then the sensor staying at the nearest corner to the coordinate position (i, j) will move to cover that point.

3.2 Total traveling Distance Based on the number of moves and individual traveling distance of the MSNs the total traveling distance, D, is calculated as:

( )[ ]( )[ ]( ) ( )1512721325

332436

−+−−++

+×−+=

nanamaa

manD

(1)

3.2.1 Total traveling Distance for different sides of varying length of the hexagons To compare our results, with the results of [3], we have considered the area of ROI as a rectangular plot of size 4500*2000 units of measure. The side of the hexagon (taken less than the sensing range) is considered 45, 50,55,60,65 units. The Total traveling distance of MSNs for their varying sensing ranges is determined. The traveling distance of the MSNs decrease as the length, a, of the side is increased. i.e., we can say that the traveling distance of the MSNs is inversely proportional to the length, a, of the side of the hexagon. The data is graphically represented in Figure 5, and in tabular form in Table 1.

Sides of regular hexagon

Figure 5. Total distance covered versus side lengths of the hexagon

Table 1 Total distance covered by MSNs for different side lengths of the hexagon

Length of side of hexagon (a)

Total traveling distance

(D) 45 690677.78 50 624499.33 55 569478.6 60 536880.36 65 465344.62

In [3] the authors have taken equilateral triangle of sides of length, a=50 units and distance traveled by the MSNs covering the rectangular region and starting at arbitrary points in the region varies from 7.38*105 to 7.54*105 which is much more than 6.91*105 obtained in our case for the regular hexagon having sides of length a=50.


44

4. Conclusions In this paper, we present a MTA for the coverage of a rectangular field by N≥4 mobile sensors. Though only 4 sensors are sufficient to cover the region, however, in case of failure, extra MSN’s kept in reserve/sleeping mode are activated. Thus the system is reliable for coverage. We also observe that as side ‘a’ of the hexagon is increased the total traveling distance covered by MSNs decreases. The hexagonal topology provides an efficient and reliable coverage as each point in the ROI is covered at least by 2-sensors.

References [1] A. Dhawan, S. K. Prasad, “Energy efficient distributed

algorithm for sensor target coverage based Performance Computing, 2008.

[2] A. Khan, C. Qiao, S.K. Tripathi, “Mobile Traversal Schemes based on Triangulation Coverage,” Mobile Netw Appl, Vol. 12, pp. 422-437, 2007.

[3] Wang, H. B. Lim, D. Ma, “A survey of movement strategies for improving network coverage in wireless sensor networks,” Computer Communications, Vol. 32, pp. 1427- 1436, 2009. [4] D. Brinza, Al. Zelikovsky, “Deeps: Deterministic energy- efficient protocol for sensor networks,” Proceedings of the International Workshop on Self Assembling Wireless Networks (SAWN), pp. 261– 266, 2006. [5] D. Tian, N. D. Georganas, “A coverage-preserving node scheduling scheme for large wireless sensor networks,” In WSN Proceedings of the 1st ACM international workshop on Wireless sensor networks

and applications, New York, NY, USA, ACM, pp. 32– 41, 2002.

[6] S. Slijepcevic, M. Potkonjak, “Power efficient organization of wireless sensor networks,” IEEE International Conference on Communications (ICC),

Vol. 2, pp. 472– 476, 2001. Authors Profile

Megha Sharma received the B.C.A and M.C.A degree from I.G.N.O.U in 2004 and 2008, respectively. She is currently working towards a Ph.D degree in computer Science at the Banasthali University of Rajasthan. Her research interests include wireless sensor networks with a focus on the coverage of

wireless sensor networks.

Prof. G. N. Purohit is a Professor in Department of Mathematics & Statistics at Banasthali University (Rajasthan). Before joining Banasthali University, he was Professor and Head of the Department of Mathematics, University of Rajasthan, Jaipur.

He had been Chief-editor of a research journal and regular reviewer of many journals. His present interest is in O.R., Discrete Mathematics and Communication networks. He has published around 40 research papers in various journals.

Mrs. (Dr.) Seema Verma obtained her M.Tech and Ph.D degree from Banasthali University in 2003 and 2006 respectively. She is a associate Professor of Electronics. Her research areas are VLSI Design, communication Networks. She has published around 18 research papers in various Journals.


45

Powers of a Graph and Associated Graph Labeling

G. N. Purohit, Seema Verma and Usha Sharma

Centre for Mathematical Sciences,

Banasthali University, Rajasthan 304022 [email protected]

Abstract: - Graph coloring is a classical topic in graph theory and vertex coloring of a graph is closely associated with channel assignment in wireless (sensor) network. Unit Disk graph is a suitable model for connectivity in sensor network. This paper is concerned with the power of a graph in general and power of Unit Disk graph in particular. L(1,1,1)- Labeling is used to avoid interference between communicating channels. We develop L(1,1,1)-Labeling of a UD graph. For this we make use of cellular partition algorithm. We have proved that cube of any UD graph can be properly colored by at most 25ω colors, where ω is the maximum clique size. Keywords: Graph Labeling, wireless network. 1. Introduction A graph G = (V, E), where V is the set of vertices and E is the set of edges. Each edge i.e. element of E is an unordered pair of element of V. Out of many induced graphs from a graph; power graph finds a special place. Powers of a graph have been considered in [1]. Square of a graph is a graph with the same vertex set in which vertices at distance 2 are connected through an edge. Cube of a graph is also the graph on the same set of vertices; however, additionally there is an edge between two vertices whenever they are at most distance 3. Graph coloring is a classical problem in graph theory and proper coloring of a graph means assigning distinct colors (labels) to adjacent vertices. The minimum number of colors required to color a graph G properly is called chromatic number of G and denoted as χ (G). A lot of research has been done on the chromatic number of graphs. χ is bounded by ω ≤ χ ≤ ∆+1 [5], where ω is the maximum clique size in the graph and ∆ is the maximum degree of graph. The chromatic number of powers of a graph has been studied in [1]. Besides proper coloring there are many types of coloring (labeling) of vertices. One such generalization is L(p,q)- labeling, in which the labels at adjacent vertices should differ by atleast p and labels at vertices at distance 2 should differ by atleast q [6]. L(p,q)- labeling problem has attracted attention of many researchers in the past [7]. Particular cases of L(p,q)- labeling (i) L(1,1)- labeling and (ii) L(2,1)- labeling have been defined and a lot of research has been done in this area. L(1,1)- labeling is also known as distance

two coloring problem and is equivalent to the proper coloring of the square of a graph, [11] includes labeling of many important graphs. Another generalization of labeling (coloring) is L(h,1,1)- labeling, in which the labels on adjacent vertices differ by atleast h and labels on vertices at distance 2 or 3 are distinct [10]. This concept is applied in channel assignment problem and in wireless (sensor) network.

Unit Disk graph [8] is another class of a graph, which finds application in modeling a wireless (sensor) network. Since the radio coverage range of sensors is based on Euclidean distance between the nodes. So we utilize the concept of Euclidean distance in a graph. This concept of Euclidean distance in a graph has given rise to a new branch termed as geometric graph theory. One can extend the concept of power of graphs to the UD graph to obtain square and cube of graphs and also Euclidean distance two graph [8] and Euclidean distance three graph. Chromatic number of UD graph and square of UD graph is considered in [8]. These results are useful in the wireless sensor network technology. In this paper we describe some powers of a graph and powers of a unit disk graph. We develop a L(1,1,1)- labeling of a UD graph by using cellular partition algorithm. This paper is organized as follows. In Section-2 we have provided some auxiliary definitions. In particular we have obtained some results related to powers of a cycle and a complete bipartite graph. In Section-3 we have defined Unit Disk graph and its powers. Some results have been proved for powers of a UD graph. In Section-4 we have given cellular partition algorithm [8]. The main result of the paper is theorem (4.1). This theorem shows that using the developed cellular partition algorithm, cube of any UD graph can be properly colored using 25ω colors, where ω is the maximum clique size. In last Section, we have given conclusion. 2. Auxiliary Definitions 2.1 Graph Powers In this section we consider different powers of graph, which finds application in channel assignments, L(p, q)- coloring of graphs etc.


46

2.1.1 Square of a graph (G2) - The Square G2 of a graph G = (V, E) is the graph, whose vertex set is V itself and there is an edge between two vertices vi and vj if and only if their graph distance (length of shortest path between vi and vj) in G is at most 2. Examples of a graph and its square graphs are given in following figures:

Figure 1. (a) Cycle C6 Figure 1.(b) Square of Cycle C6

2.1.2 Cube of a graph (G3) - The Cube G3 of a graph G (V, E) is the graph, whose vertex set is V and there is an edge between two vertices vi and vj if and only if their graph distance in G at most 3. Examples of a graph and its cube graphs are given in following figures:

Figure 2. (a) Cycle C6 Figure 2. (b) Cube of Cycle C6 We can generalize the above definitions as follows:- 2.1.3 Kth power of a graph (Gk) - The kth power Gk of a graph G (V, E) is the graph, whose vertex set is V and there is an edge between two vertices vi and vj if and only if their graph distance in G is at most k. As a special case we prove the following results for cycle Cn and complete bipartite graph Km,n. Theorem 2.1 If G = Cycle with n vertices (Cn), Then

Proof: Case (i): Let G = Cn (4≤n≤2k) and let vi and vj be any two arbitrary vertices of G. The maximum distance between vi and vj could be k in this case and thus any pair of vertices could have a graph distance at most k. Thus max {d (vi, vj)│∀ vi, vj є V}≤k.

From the definition of Gk, two vertices will be adjacent if d

(vi, vj) ≤ k. Since this condition is satisfied by all pairs of vertices in G. Therefore all pairs of vertices will be adjacent

in Gk and hence Gk will be a complete graph on n vertices and thus (C n)k =Kn. Case (ii): Let G = Cn (n≥2k+1) and let vi be an arbitrary vertex of graph G. We have to show that deg (vi) = 2k ∀ vi є C n. We know that deg (vi) = 2, ∀ vi є C n. From the definition of Gk

, two vertices will be adjacent if the distance between them is at most k. There are exactly 2k vertices which are at most at a distance k from vi. On one side of vi these k vertices vi+1, vi+2, vi+3………. vi+k are at distance 1, 2, 3…….k from vi

respectively. Similarly on the other side of vi these k vertices vi-1, vi-2, vi-3………. vi-k are at distance 1, 2, 3…….k from vi

respectively. Out of these 2k vertices vi+1 and vi-1 are already adjacent to vi in Cn and remaining 2k-2 vertices will be made adjacent to vi in (Cn) k. Therefore deg (vi) = 2+2k-2 = 2k. Thus (Cn) k will be a 2k-Regular graph on n vertices.

Theorem 2.2 If G = Km,n, then Gk = Gk-1= Gk-2 =……………= G3 = G2 = Km+n. Proof: Let G = Km,n be a bipartite graph. Let V1 and V2 be two partitions of vertex set V of G with m and n number of vertices respectively. Let vi be an arbitrary vertex of V1. Then all the vertices of V2 are at distance 1 from vi. Moreover all other vertices of V1 are at distance 2 from vi. Since vi is an arbitrary vertex therefore this is true for all vi’s in V1 as well as for all vertices in V2. Thus all the pairs of vertices are adjacent in G2. Thus G2 will be a complete graph on m+n vertices. Gk (k>2) will not change Km+n. Thus Gk = Gk-1= Gk-2 =……………= G3 = G2 = Km+n.

2.2 Labeling of a graph G (V, E)

2.2.1 L (p, q) – Labeling - For two positive integers p and q, an L(p,q)- Labeling of a graph G is a function C:V(G) → N such that C(vi) - C(vj) ≥ p if vertex vi and

vj are adjacent and C(vi) - C(vj) ≥ q if vertex vi and vj

are at distance 2.

In particular L(1, 1)- labeling and L(2,1)- labeling are well known examples of L(p,q)- Labeling.

2.2.2 L (1,1) – Labeling - It is also called the proper labeling of a graph G. It is the labeling of the vertices with non negative integers such that the labels on adjacent vertices differ by at least 1.

2.2.3 L (2,1) – Labeling - It is a labeling of the vertices with non negative integers such that the labels on adjacent vertices differ by at least 2 and the labels on vertices at distance 2 differ by at least 1.

We can generalize the above definition as follows:-


47

2.2.4 L (p,q,r) – Labeling - For three positive integers p, q and r, an L(p,q, r)- Labeling of a graph G is a function C:V(G) → N such that C(vi) - C(vj) ≥ p if vertex vi and

vj are adjacent, C(vi) - C(vj) ≥ q if vertex vi and vj are at

distance 2 and C(vi) - C(vj) ≥ r if vertex vi and vj are at distance 3. In particular L(1,1,1)- Labeling is more useful in channel assignment problem and in wireless (sensor) network than the others. 2.2.5 L (1,1,1) – Labeling - It is the labeling of the vertices with non-negative integers such that the labels on adjacent vertices, on vertices at distance 2 and 3 are different.

We can generalize it as follows:-

2.2.6 L (d1,d2,d3,…..di,…..dk) – Labeling:- It is a labeling of the vertices with non-negative integers such that the labels on vertices at distance i from each other differ at least by di. 3. Powers of a Unit Disk graph For the sake of completeness, we first define unit disk graph. 3.1 Unit Disk Graph- A graph G is a Unit Disk graph if there is an assignment of unit disks centered at its vertices such two vertices are adjacent if and only if one vertex is within the unit disk centered at the other vertex. We denote a unit disk graph by GUD.

3.2 Square of a Unit Disk Graph (GUD2) - The Square

GUD2

of a Unit Disk graph GUD (V, E) is the graph whose vertex set is V and there is an edge between two vertices vi and vj if and only if their graph distance in GUD is at most 2.

3.3 Euclidean distance two graph of a Unit Disk graph (GUD

ED2) - Euclidean distance two graph of a unit disk graph GUD (V, E) is the graph whose vertex set is V and there is an edge between two vertices vi and vj if and only if their Euclidean distance in GUD is at most 2.

3.4 Cube of a Unit Disk graph (GUD3) - The cube GUD

3 of a Unit Disk graph GUD (V, E) is the graph whose vertex set is

V and there is an edge between two vertices vi and vj if and only if their graph distance in GUD is at most 3.

3.5 Euclidean distance three graph of a Unit Disk graph (GUD

ED3) - Euclidean distance three graph of a unit disk graph GUD (V, E) is the graph whose vertex set is V and there is an edge between two vertices vi and vj if and only if their Euclidean distance in GUD is at most 3.

Figure 4. (a): Cube of a UD graph (GUD

3) Figure 4. (b): ED-3 graph of a UD graph (GUD

ED3) Now we discuss some results relating to GUD and GUD

ED.

Theorem 3.1 For any Unit disk graph GUD, GUD2 ⊆ GUD

ED2.

Proof: The proof of this theorem is given in [8].

Theorem 3.2 For any Unit disk graph GUD, GUD3 ⊆ GUD

ED3.

Proof: Let GUD be a Unit Disk graph. GUD3 be the cube of

GUD and GUDED3 be the Euclidean distance three graph of

GUD. Since both the graph are on the same vertex set. So it is sufficient to prove this theorem that edge set of GUD

3 is the subset of edge set of GUD

ED3. Let (c, w) be an edge in GUD3.

There must exist two vertices u & v such that (c, u), (u, v),

Figure 3. (a): Square of a UD graph

Figure 3. (b): ED-2 graph of a UD graph

4(a)

4(b)


48

(v, w) are three edges in GUD. Since GUD is a Unit Disk graph.

If we consider dEd(c, w) denotes the Euclidean distance between c and w, then

dED(c, w) ≤ dED(c, v) + dED (v, w)

≤dED(c, u) + dED (u, v) + dED (v, w)

= 1+1+1 = 3.

Hence dED (c, w) ≤ 3. Thus the edge (c, w) is an edge in GUD

ED3.

Hence GUD3 ⊆ GUD

ED3

Further GUD3 may be proper subgraph of GUD

ED3 in some instances. We will show that there might be an edge in GUD

ED3 but not in GUD3. As shown in figure 4(b), there might

be a vertex x in GUDED3 such that 1<dED(c, x) ≤ 3 and but

there are no two vertices u' and v' such that (c, u'), (u', v') and (v', x) are edges in GUD. Thus (c, x) is an edge in GUD

ED3 but not in GUD

3. Similarly there might be a vertex y in GUD

ED3 such that 1<dED(c, y) ≤ 2 but there are no vertex w' such that (c, w') and (w', y) are edges in GUD. Thus (c, y) is an edge in GUD

ED3 but not in GUD3.

Theorem 3.3 For any UD graph GUD, a coloring scheme χ (GUD

ED3) for coloring GUDED3 would also color GUD

3 which is equivalent to L(1,1,1)- labeling of GUD.

Proof: Since we’ve proved in previous theorem, any GUD3 be

a subgraph of GUDED3 then ∃ a coloring scheme χ (GUD

ED3) to color GUD

ED3 properly could be sufficient to any of its subgraph. Therefore it would also color GUD

3. Since L(1,1,1) labeling of GUD is equivalent to proper coloring of GUD

3. Thus χ (GUD

ED3) fulfill L(1,1,1) labeling of GUD.

4. Cellular Partition Algorithm

The concept of UD graph as well as labeling can be applied in wireless sensor networks since we can model a wireless sensor network as a UD graph. In this modeling sensors are denoted as vertices. The sensing coverage area of a sensor is represented by a unit disk centered at the corresponding vertex. The connectivity between two sensors is determined as if one sensor is within the sensing coverage area of another sensor. If GUD represent a model of a wireless sensor network, then GUD

2 and GUD3 provide possible interfering

sensor nodes. To avoid this interference we need a proper labeling for GUD

3 which is equivalent to L(1, 1, 1)- Labeling of GUD.

In order to cover the targeted area by sensors, we have to divide the whole area in smaller cells (area). We have chosen regular hexagons to cover the whole plane based on the observation that hexagon is the most suitable polygon which could cover the plane efficiently. It is the most suitable tile that could cover the plane with no overlap and thus it is the most efficient way to cover the plane.

Now in order to label the nodes we adopt the Cellular Partition algorithm. In this algorithm first of all we partition the whole plane in unit hexagonal cells with a side length ½ thus the diagonal length of each cell is 1. If there is any UD graph in this plane, vertices of the graph inside the cell will form a clique, since no two vertices in the same hexagon have a Euclidean distance greater than 1. Let the maximum clique size be ω then there can not be more than ω vertices in the same hexagonal cell. Since we know that ω colors are sufficient to color each hexagonal cell. Therefore we can color the whole graph properly.

Using the above Cellular Partition algorithm we prove the following theorem:

Theorem 4.1 Euclidean distance three graph GUDED3 for any

UD graph GUD can be properly colored by at the most 25ω colors where ω is the maximum clique size.

Proof: We partition the whole plane into hexagonal cells with side ½ and diagonal 1. All vertices included in any hexagon would form a clique. Since ω is the maximum clique size, so we could place at most ω vertices into each cell.

Next we construct a patch of 25 hexagons* and use 25ω color to color the patch. An example of the patch is shown in figure [5] and keeps the same orientation of patches of 25 to cover the whole plane as shown in figure [6]. Now we prove that a vertex in ith hexagon in a patch would be at a Euclidean distance of at least 3 to any other vertex in the ith

hexagon in any other adjacent patch.

We maintain the same numbering orientation in a patch for all patches in the whole plane. So the distance between two vertices in ith hexagons in adjacent patches is constant. As an example let A, B and C be the centers of centre hexagons in the three adjacent patches as shown in figure [6]. Their distance can be computed as: We know that AB = 10* 3 /4 = 4.33 > 4. Also we have AD = ½ + ½ + 1+ ½ + 1+ ¼ = 15/4

Figure 5. A patch of 25 hexagons


49

Figure 6. Cover the whole plane with the patches of 25 hexagons

CD = 5* 3 /4 AD2 + CD2 = AC2

Therefore AC = 22

43*5

415

+

= 4.33 > 4. Since the

vertices in both the hexagons are at distance ½ only from their centers A, B and C then the distance from any vertex in hexagon with center A to any vertex in hexagon with center B will be greater than 3. Similarly the distance from any vertex in hexagon with center A to any vertex in hexagon with center C will also be greater than 3. Therefore our patch fulfills the coloring of GUD

ED3.

In the above coloring scheme each hexagon with a certain color pattern is far away from its sibling hexagon with same color pattern. So it is a valid coloring for GUD

ED3. We can properly color any GUD

ED3 graph with 25ω colors. Using Theorem 3.2, it follows that it is valid color scheme for GUD

3 also.

5. Conclusion

Using the developed cellular partition algorithm, cube of any UD graph can be properly colored using 25ω colors, where ω is the maximum clique size. This is equivalent to L(1,1,1)- labeling of unit disk graph and can be used to avoid interference between communicating channels in wireless (sensor) network. The number 25ω is the upper bound and we are looking for obtaining a suitable lower bound also.

Acknowledgement

Ms USHA SHARMA, one of the authors of this paper acknowledges the grant received from Department of Science & Technology (D.S.T.), Government of India, New Delhi for carrying out this research.

References

[1] N. Alon and B. Mohar, “The chromatic number of graph

powers”, Comb. Probab. Comput. 11, 1 (2002), 1–10. [2] G. J. Chang and D. Kuo, “The l(2,1)-labeling problem

on graphs”, SIAM J. Discret. Math. 9, 2 (1996), 309–316.

[3] B. N. Clark, C. J. Colbourn and D. S. Johnson, “Unit disk graphs”, Discrete Math. 86, 1-3 (1990), 165–177.

[4] A. Graf, M. Stumpf and G.Weisenfels, “On coloring unit disk graphs”, Algorithmica 20, 3 (1998), 277–293.

[5] D. B. West, “Introduction to Graph Theory”, Second edition Prentice Hall, 2001.

[6] B. M. K. Q. Peter Bella, Daniel Kral, “Labeling planar graphs with a condition at distance two,” in Proceedings 2005 European Conference on Combinatorics, Graph Theory and Applications, 2005.

[7] M. Hall d´orsson, “Approximating the l(h, k)-labelling problem,” Engineering Research Institute, University of Iceland Technical, Tech. Rep. Report No. VHI 03-2005, Available: citeseer.ist.psu.edu/252952.html

[8] T. Ren, K. L. Bryan, and L. Thoma, “On coloring the square of unit disk graph,” University of Rhode Island Dept. of Computer Science and Statistics, Tech. Rep., 2006.

[9] Kevin L. Bryan, Tiegeng Ren, Lisa DiPippo, Timothy Henry, Victor Fay-Wolfe, “Towards Optimal TDMA Frame Size in Wireless Sensor Networks”, University of Rhode Island Dept. of Computer Science and Statistics, Tech. Rep.

[10] T. Calamoneri, E.G. Fusco, R.B. Tan and P. Vocca, “L (h,1,1)- labeling of outerplanar graphs”, Mathematical Methods of Operations research, Volume 69, Number 2, May 2009, 307-312.

[11] T. Calamoneri, “The L(h, k)-Labelling Problem: A Survey and Annotated Bibliography”, The Computer Journal Vol. 49 No. 5, 2006.


50

A Method of Access in a Dynamic Context Aware Role Based Access Control Model for Wireless

Networks 1Dr. A.K. Santra, 2Nagarajan S

1Director and Professor, MCA Department, Bannari Amman Institute of Technology, Sathayamangalam, TamilNadu. 2Research Scholar, Bharathiar University, Coimbatore and Selection Grade Lecturer, Alliance Business Academy, Bangalore, Karnataka.

Abstract: This paper address security in dynamic context aware systems. Context awareness is a emerging as an important element in wireless systems. Security challenges in context aware systems include integrity, confidentiality and availability of context information as well as end user’s privacy. The paper addresses the dynamic changes happening in the mapping between the roles and permissions depending on context information. The paper presents a access control method using artificial neural networks. It represents the data in terms of bits to express the roles and permissions which helps in reducing the data transmission and is a good fit for wireless networks with lower bandwidth. It also introduces a novel method for storing the information in a reduced format. Instead of accessing the access control tables the machine is learning it, which in turn reduces the time required to access the tables. Being dynamic in nature there is no requirement for changes, any change is taken care by the machine learning itself. Further, the algorithm is simple and easy to implement in wireless networks. Keywords: Dynamic Context, Wireless Networks.

1. Introduction

It has been proved that Dynamic Role Based Access Control can manage Access Control and security, more and more mobile devices are incorporating this feature. Pervasive communication technology is becoming a everyday feature and it is changing the way of communicating with the external world. This type of DRBAC requires the following tables: 1. User Location Table 2. User Role Table 3. Role – Permission Table and 4. Mutual Exclusive role table. Each time anybody accesses the system the first three tables are searched. Further, there is a very complex mapping of Location, users, roles and permissions. It has been observed that frequently searching the tables reduces the efficiency of access control. An disadvantage of wireless devices are that they have less power, storage, computing and transmission abilities. Hence, performing access control in wireless environments is actually more complex than that I wired environments. Therefore, any approach to access control must be relatively simple and very efficient. This paper addresses the following points: It gives a access control algorithm and storage is reduced using the EAR decomposition and is retrieved accordingly. It also uses a ANN to train the system so that this procedure is learnt by the system, rather than searching the tables. This algorithm assigns the user with different permissions in different sessions depending on the context aware data

available at that point of time. This reduces the data storage and transmission for using only the bits making it very much easy to complement in networks where the bandwidth of the network is very low. The anytime, anywhere access infrastructures is to enable a new generation of applications that can leverage continuously manage, adapt and finally optimization is required. The major challenge faced in Wireless applications is managing the security of the system using Access Control Lists. ACL's is a very common mechanism used in Access Control. It has been observed that the ACL's are used to check for permission to access resources or services. Another point to be noted at this juncture is such type of approach is very inadequate for wireless applications, since most proposed models do not take care of context information into consideration. There is a need for giving control in a dynamic way as the context changes according to location, time, system resources, network security configuration etc., Therefore, access control mechanism that changes the permission of a user dynamically based on context information is very much essential. In this direction [3] have proposed a GRBAC Model and representing the system using State Machines. Using this model, It is representing the information for the new algorithm proposed and show how it can be stored and retrieved. Then finally, show how this can be used to train the system without accessing the matrix.

2. Background

Location, User, Role and Permission are the major components of a DRBAC which are represented as follows: L = {L1, L2, ........................Li} U = {U1, U2, .......................Ui} R = {R1, R2, .......................Ri} P = {P1, P2, ........................Pi} T = {T1, T2, T3} The permission only directly maps to one role. In case many roles want to own the same permission, this need to be done using role inheritance. Since conflicted permissions also needs to be addressed.


51

3. Dynamic Context Aware Role Based Access Control

DRBAC addresses the dynamic requirement of applications in pervasive environments. It extends the traditional RBAC model to use dynamic context information while making access control decisions. The DRBAC addresses the following:

1. A user's access privileges must change when the user's context changes.

2. A resource must adjust its access permission when its system information changes.

4. DRBAC Definitions

The DRBAC definitions are taken from the RBAC formalisms presented in [3] and [4] USER: A user is an entity whose access is being controlled. USERS represents a set of users. ROLES: A role is a job function within the context of an organization with some associated semantics regarding the authority and responsibility conferred on the user assigned to the role. ROLES represents a set of roles. PERMS: A permission is an approval to access one or more RBAC protected resources. PERMS represents a set of permissions. LOCATIONS: Locations is the set of points from where the user accesses the resources. LOCATIONS is the set of points of access. TIMES: Times is the time at which the user access the resources. Times is the set of time at which the user has the access. SESSIONS: A session is a set of interactions between subjects and objects. A user is assigned a set of roles during each session. The active role will be changes dynamically among the assigned roles for each interaction. SESSIONS represents a set of sessions. UA : UA is the mapping that assigns a role to a user. In the session, each user is assigned a set of roles, the context information is used to decide which role is active. The user will access the resource with the active role. PA : PA is the mapping that assigns permissions to a role. Every role that has a privilege to access the resource is assigned a set of permissions, and the context information is used to decide which permission is active for that role. Definition of the Agent: A Central Authority checks for the user's access rights. And gives the privileges that are active for him in that session.

5. Explanation of the DRBAC Model

The environment considered is an educational institute. The designations are Professor, Associate Professor, Assistant Professor and Teaching Assistant. At office they will have both read and write permissions. For this we represent the locations, roles and Time in the following way: Locations

L1 = Campuses Abroad L2 = Campuses coming under the home country L3 = Campuses in each City L4 = Campuses within the city L5 = Residence Time T1 = 8:00 AM to 8:00 PM (Office Hours) T2 = 5:30 AM to 7:59 AM (Morning) T3 = 8:01 PM to 5:29 AM (Night) Roles For Time T1 R1 = Professor R2 = Associate Professor R3 = Assistant Professor R4 = Teaching Assistant R5 = Professor Remote R6 = Associate Professor Remote R7 = Assistant Professor Remote R8 = Teaching Assistant Remote For Time T2 R9 = Professor R10 = Associate Professor R11 = Assistant Professor R12 = Teaching Assistant R13 = Professor Remote R14 = Associate Professor Remote R15 = Assistant Professor Remote R16 = Teaching Assistant Remote For Time T3 R17 = Professor R18 = Associate Professor R19 = Assistant Professor R20 = Teaching Assistant R21 = Professor Remote R22 = Associate Professor Remote R23 = Assistant Professor Remote R24 = Teaching Assistant Remote Permission P1 = Append P2 = Create. P3 = Execute. P4 = Get attribute. P5 = I/O Control. P6 = Link. P7 = Lock. P8 = Read. P9 = Rename. P10 = Unlink. P11 == Write. The Access Control Algorithm for wireless applications. For the sake of this study it is considered that static IP addresses are used. The wireless infrastructure implementing a WLAN is used for the logins inside the campus; While Broadband wireless internet is used to login remotely. Step 1: Using IPSec Labeling the process of authentication is done as described in [5]. Step 2: Using the IP address associated with the user the location of the user is determined. Step 3: Depending on the user's location a role is assigned which is further associated with permissions.


52

Using the following information we try to ascertain whether a user is permitted to login from a particular location or not using matrix1. If the said user has access rights from that location the step 2 of the algorithm is executed i.e., is mapping the IP address to a role else the access right is denied. Matrix1

L1 L2 L3 L4 L5 U1 1 1 1 1 1 U2 0 1 1 1 1 U3 0 1 1 1 1 U4 0 0 0 1 0 . 0 0 0 1 0

U5 1 1 1 1 1 The function of the second matrix defines the relationship between the Location and roles for the time the user logs in. Depending on the time the user logs in the roles are assigned. This is used to check whether a role has access rights at various locations are not. Further, the permission for the roles are defined during the time the role is created. If the role column in the matrix is 1 it means that role can be provided access for that location and further step 3 of the algorithm is executed else the access to that role is denied. Matrix2 For Time T1

R1 R2 R3 R4 R5 R6 R7 R8

L1 1 0 0 0 0 0 0 0

L2 1 1 1 0 0 0 0 0

L3 1 1 1 0 0 0 0 0

L4 1 1 1 1 0 0 0 0

L5 1 1 1 0 1 1 1 0 For Time T2

R9 R10 R11 R12 R13 R14 R15 R16 L1 1 0 0 0 0 0 0 0 L2 1 1 1 0 0 0 0 0 L3 1 1 1 0 0 0 0 0 L4 1 1 1 1 0 0 0 0 L5 1 1 1 0 1 1 1 0

For Time T3

R17 R18 R19 R20 R21 R22 R23 R24

L1 1 0 0 0 0 0 0 0

L2 1 1 1 0 0 0 0 0

L3 1 1 1 0 0 0 0 0

L4 1 1 1 1 0 0 0 0

L5 1 1 1 0 1 1 1 0

Based on the permission rights for that user the access is allowed. These two matrix are represented in the form of a graph and then use the open ear decomposition technique to reduce this information and store it.

6. Performance test of the algorithm

The test bed was created as a kernel program in SeLinux. It is allowed to run with the same modules that Se Linux has in addition to the modules created for this purpose. Whenever somebody logins into the system it uses the authentication methods presently provided by the operating system. Using this to our advantage we put our static addresses specific to the location based on the labeling of IPSec object called labeled IPSec. This particular feature is available in mainline Linux version 2.6.16 itself. This does the authorization process as described in [5] and also we use the same information to determine the location of the user. Once the user's location is ascertained the next step is to look out for the time at which this login has been requested. This is done with the help of the system clock. With this context information that is generated, access roles are accordingly assigned. The SELinux user identities are different from UNIX identities. Here, for experimentation the normal roles defined are R1, R2, R3, R4, ............R24 and the corresponding Selinux roles defined are R1_r, R2_r, R3_r, R4_r, ......R24_r. These roles are associated with the user. The normal user are U1, U2, U3, U4, ........Un and the corresponding Selinux users defined are U1_u, U2_u, U3_u, U4_u, ........... Un_u. Here _r identifies the roles while _u identifies the user. SELinux user identities are different from UNIX identities. They are applied as part of the security label and can be changed in real time under limited conditions. SELinux identities are not primarily used in the targeted policy. In the targeted policy, processes and objects are system_u, and the default for Linux users is user_u. When identities are part of the policy scheme, they are usually identical to the Linux account name (UID), and are compiled into the policy. In such a strict policy, some system accounts may run under a generic, unprivileged user_u identity, while other accounts have direct identities in the policy database _t identifies type. SELINUX_SRC/rbac is the place in which roles are allowed to attain which other roles. Types are the primary security attribute Selinux uses in making authorization decisions as defined in permissions above. This is defined in /etc/security/selinux/src/policy. Depending on this roles can be assigned. 7. Representation of the Matrix and decomposition / retrieval Using the three Matrix defined in the above method, the next step is to apply the well known Hungarian Algorithm to represent the matrix in the form of a graph. The Steps in the Hungarian Algorithm is as follows:


53

Step 1 Generate initial labeling L and matching M in EL. Step 2 If M perfect, stop. Otherwise pick free vertex U such that it belongs to X. Set S = { U } , T = Null. Step 3

If NL (S) = T, Update labels (forcing NL(S) ≠ T)

αl = mins € S, y does not belong to T. l(v) – αl if v € S l’(v) = l(v) + αl if v € T l(v) otherwise Step 4 If Nl (S) ≠ T, Pick y € Nl (S) – T If y free, u – y is the augmenting path, Then Augment M and Go to step 2. Else If y matched, say to z, extend alternating tree: Such that, S = Su { z }, T = T U { y } Go to step 3. Matrix1 and its graph representation G1

Matrix2 and its graph representation G2

Similarly, the graphs for the other two matrix is drawn and reduced as shown. Now, using the two graphs we apply the path ear decomposition algorithm. The steps of the path Ear decomposition algorithm is as follows: An ear decomposition D = [ P0 , P1 , P2 , ………., Pr-1 ] of an undirected graph G = (V, E) is a partition of E into an ordered collection of edge-disjoint simple paths P0 , P1 , P2 , ………., Pr-1 such that P0 is an edge, P0 U P1 is a simple cycle, and each end point of Pi , for i > 1, is contained in some Pj , j < i, and none of the internal vertices of Pj are contained in any Pj , j < i. The paths in D are called ears. An ear is open if it is non-cyclic and is closed otherwise. A trivial ear is an ear containing a single edge. D is an open ear decomposition if all ears are open. Let D = [ P0 , P1 , P2 , ………., Pr-1 ] be an ear decomposition for a graph G = (V, E). For a vertex v in V, we denote by ear(v), the index of the lowest numbered ear that contains v; for an edge e = (x,y) in E, we denote by ear(e) (or ear(x,y)), the index of the unique ear that contains e. A vertex v belongs to Pear(v). The path ear decomposition algorithm: Input: A connected graph G = (V, E) with a root r € V, and with V = n. Output : A depth first search tree of G, together with a label on each edge in E, indicating its ear number. Set T of edges; integer count; Procedure df s(vertex v); { * This is a recursive procedure. The call df s(v) of the main program constructs a depth first search tree T of G rooted at r; the recursive call df s(w) constructs the sub tree of T rooted at w. The depth first search tree is constructed by placing the tree edges in the set T and labeling the vertices in the sub tree rooted at vertex v in pre-order numbering, starting with count. The procedure assigns ear labels to the edges of G while constructing the depth first search tree. An edge that does not belong to any ear is given the label (∞, ∞). Initially, all vertices are unmarked. * } Vertex w; ‘mark’ v; Pre-order(v) := count; count := count + 1; low(v) := n; ear(v) := (n,n); For each vertex w adjacent to v { * This for loop performs a depth forth search of each child of v in turn and assigns ear labels to the tree and non tree edges incident on vertices in the sub trees rooted at the children of v. * } If w is not marked Add (v,w) to T; parent(w) : = v; df s(w); If low(w) ≥ pre-order(w) ear(parent(w), w) := (∞, ∞) Low(w) < pre-order(w) ear(parent(w),w) := ear(w) Fi; Low(v) := min(low(v), low(w)); Ear(v) := lexmin(ear(v), ear(w)) If w is marked If w ≠ parent (v) Low(v) := min(low(v), pre-order(w));


54

Ear (w, v):= (pre-order(w), pre-order(v)) Ear (v) := lexmin(ear(v), ear(w,v)); Fi Fi Rof End df s; {* Main program *} T: = Null set; count: = 0; df s(r); Sort the ear labels of the edges in lexicographically non-decreasing order and relabel distinct labels (expect labels (∞, ∞)) in a order as 1,2,3,,4,………; Relabel the non tree edge with label 1 as 0 End. Using the algorithm the graph G1, G2 and G3 reduces to G11, G21 and G31 respectively. Graph G1 reduced to the form G11 P1 = { < U1, L1 > < U5, L1 > } P2 = { < U1, L2 > < U5, L2> } P3 = { < U1, L3 > < U5, L3 >} P4 = { < U1, L4 > < U5, L4 >} P5 = { < U1, L5 > < U5, L5} P6 = { < U2, L2 > U3, L2 > < U3, L3 > } P7 = { < U2, L3 > } P8 = { < U3, L4 > < U4, L4 > } P9 = { < U3, L5 > } P10 = { <U4, L4 > } Therefore, G11 = { P1, P2, P3, P4, P5, P6, P7, P8, P9, P10} Graph G2 reduced to the form G21 P1 = { < L1, R1> < L2, R1 > < L2, R2 > < L3, R3 > < L4, R3 > < L4, R4 >} P2 = { < L2, R3 > < L5, R3 > < L5, R5 > } P3 = { < L5, R6 > < L5, R7 > } Therefore, G21 = { P1, P2, P3} Similar operation is performed on the other two graphs. G11 and G21 are referred to as the partition matrix and can be called partition path matrix. The path decomposition is edge disjoint one Whence the union of the path reduced will give the entire graph G1 and G2.

8. Conclusion

It has been observed that any dynamic context aware system needs to search relative tables to get the user permissions. This paper is presenting a dynamic context aware algorithm using SElinux where the number of tables are reduced. It also shows a way to store it and retrieve. Executing our module the roles are assigned according to the location and time. Hence it can be implemented with ease in a wireless networked environment.

Acknowledgements

We Would like to thank Prof. K. A Venkatesh, HOD Department of Computer Applications, Alliance Business Academy for all his support and discussions. We would also like to thank Mr. Mahesh M S for the experimental support provided in the lab during the preparation of this algorithm and module.

References [1] Efficient Access Control in Wireless Networks, Kun

Wang, Zhenguo Ding, Lihua Zhou, Proceedings of IEEE/WIC/ACM International Conference on Web Intelligence and Intelligent Agent Technology. 85-88, ISBN:0-7695-2749-3, 2006.

[2] Fast Access Control algorithm in Wireless Network, Kun Wang, Zhixin Ma, This paper appears in: Grid and Pervasive Computing Workshops, 2008. GPC Workshops '08. The 3rd International Conference on, ISBN 978-0-7695-3177-9, 347p – 351p, 25-28 May 2008.

[3] Context-Aware Dynamic Access Control for Pervasive Applications,. G. Zhang and M. Parashar, Proceedings of the Communication Networks and Distributed Systems Modeling and Simulation Conference (CNDS 2004), 2004 Western MultiConference (WMC), pp. 219 . 225, January 2004.

[4] Supporting relationships in access control using role based access control. K. Beznosov, J Barkley and J Uppal, Symposium on Access Control Models and Technologies, Proceedings of the fourth ACM workshop on Role-based access control, Fairfax, Virginia, United States, 55p – 65p, ISBN:1-58113-180-1 1999.

[5] Leveraging IPsec for Distributed Authorization, Trent Jaeger, David King, Kevin Butler, Jonathan McCune, Ramon Caceres, Serge Hallyn, Joy Latten, Reiner Sailer and Xiolan Zhang. nsrc.cse.psu.edu/tech_report/NAS-TR-0037-2006.pdf, 2006

Authors Profile Dr. A.K.Santra is presently working as the Director (Computer Applications), at the Bannari Amman Institute of Technology in Sathyamangalam. He has close to 40 years of experience both in the industry and Teaching. He published 17 papers in various International

Journals and conferences. He is presently guiding a number of students for their Ph. D. degrees. He is on the board and a reviewer in various International Journals.

Mr. Nagarajan S is presently working as Selection Grade Lecturer, at the Alliance Business Academy, Bangalore. He is also a Research Scholar at Bharathiar University at Coimbatore. He has nearly about 13 years of Industry and teaching experience. He has published one international paper in an

International Journal and 5 in various conferences.


55

Design of a Novel Cryptographic Algorithm using Genetic Functions

Praneeth Kumar G1 and Vishnu Murthy G2

1 C V S R College Of Engineering, Ghatkesar, Andhra Pradesh, India

[email protected]

2C V S R College Of Engineering, Ghatkesar, Andhra Pradesh, India

[email protected]

Abstract: Information Security plays a key role in the field of modern computing. Here, at this paper we present a new cryptographic algorithm which is proven to be resistant to Cryptanalysis, Bruteforce and timing attacks. As the algorithm uses Blum Blum Shub Genrator, A Cryptographically Secure Pseudorandom Bit Generator (CSPRBG) for deriving the key and Gentic Funtions in the process of Encryption. A comparison of the proposed technique with existing and industrially accepted RSA and Triple-DES has also been done in terms of resistance to attacks and the various features of the algorithm.

Keywords: Encryption, Decryption, Blum, Blum, Shub Generator, Genetic Functions.

1. Introduction Information Security plays a vital aspect of modern

computing systems. With the global acceptance of the Internet, virtually every computer is connected to every other. So at this point of time maintaining secrecy and security of information has become necessity. For these reasons different types of research works on encryption and decyption is going on so that various algorithms are developed in this field. The process of encoding a message so that it can be read only by the sender and the intended recipient is known as encryption. The encoded version is known as cipher text and process of decoding the cipher text is known as Decryption.

The Algorithm uses Blum Blum Shub Generator for generating key and Genetic functions “CROSSOVER” and “MUTATION” in the process of encryption and decryption. The Algorithm uses a key of four parameters, for security which makes it resistant against Bruteforce attack.

Key = {p, q, s, k} Where, p, q are two large prime numbers and s is a randomly chosen number where s is relatively prime to n (product of p and q) and k is Key Size used.

2. Literature Survey

2.1 Blum, Blum, Shub Generator A popular approach for generating secure pseudorandom number is known as the Blum, Blum, Shub (BBS) generator, named for its developers[1]. The procedure is as follows. First, choose two large prime numbers, p and q ,

that both have remainder of 3 when divided by 4. That is, p≡q≡3 (mod 4)

means that (p mod 4)= (q mod 4)= 3. Let n = p X q. Next, choose a random number s, such that s is relatively prime to n; this is equivalent to saying that neither p nor q is factor of s. Then the BBS generator produces a sequence of numbers Xi according to the following algorithm: X0 = s2 mod n. for i =1 to infinite Xi = (Xi-1)2 mod n The BBS is referred to as a cryptographically secure pseudorandom bit generator (CSPRBG). A CSPRBG is defined as one that passes the next- bit test , Which is defined as follows: “A Pseudo random bit generator is said to pass the next-bit test if there is not a polynomial-time algorithm that, an on input of the first k bits of an output sequence, can predict the (k+1)st bit with probability significantly greater than 1/2”. The security of BBS is based on the difficulty of factoring n. That is, we need to determine its two prime factors p and q. 2.2 Genetic Functions In the proposed algorithm we use two genetic functions “CROSSOVER” and “MUTATION”.

Crossover is a genetic function which can be described by the following figure: As Illustrated in the figure the Binary representation of key and plain text are Crossected. We have two forms of crossover: Single and Double Crossover. Taking 1 breaking point for a single crossover and 2 breaking points for double crossover. Crossover :

Mutation is a genetic function where the bit at a given position is inversed (i.e., 0 to 1 and vice versa).


56

3. Proposed Algorithm The algorithm consists of two phases where the first phase

is of generating random numbers and the other performs encryption/ decryption.

3.4 Key Generation The algorithm uses a 4-tuple key {p, q, s, k} where p and q are large prime numbers, s is a chosen random number which is relatively prime to n, the product of p and q and k, the key size. The key size is of Variable one.

Then, the algorithm uses the Blum, Blum, Shub Generator for generating the random numbers (Which is described in Section 2.1) which are used as keys in each iteration for encryption. 1. Choose p=7 and q=19 2. Implies, n= 7 X 19 = 133 3. Choose s=100, relatively prime with 133 4. Then, X0=s2mod n= (100)2 mod 133= 25 X1=(X0)2 mod n= (25)2 mod 133= 93 X2=(X1)2 mod n= (93)2 mod 133= 4 X3=(X2)2 mod n= (4)2 mod 133= 16 . . . . Here, the key is represented as {7, 19, 100, 8 }.

3.5 Encryption/ Decryption Algorithm The proposed algorithm follows the below given method

for encryption and decryption. The Random numbers should be generated concurrently in both the processes.

3.2.4 Encryption The Encryption process is carried out as : for every bit in the file until EOF

if random number generated is odd perform CROSSOVER between plain text(binary representation of ASCII value) and the random number(in binary representations ) where the breaking point is xi%k .

else if the number generated is even perform Double CROSSOVER between plain text(binary representation of ASCII value) and the random number(in binary represenations) where the first breaking point is xi%k and second one is (xi+s)%k.

perform MUTATION at the (2*xi)%k position in the offsprings..

The set of two numbers from the above output is the cipher text. Single Crossover Suppose that the Message is AB and Key is CD

Where, A is part of Plain text before breaking point B is part of Plain text after breaking point. C is part of Key before breaking point. D is part of Key after breaking point.

A B A D A’ D è(Crossover) è(Mutation) C D C B C’ B

Double Crossover Suppose that the Message is AB and Key is CD Where, A is part of Plain text before first breaking point

B is part of Plain text between first and second breaking points. C is part of Plain text after the second breaking point.

D is part of Key before first breaking point. E is part of Key between first and second breaking point. F is part of Key after the second breaking point.

A B C D B F D B’ F è(Double Crossover) è(Mutation) D E F A E C A E’ C Then, the Plain Text’s (Binary representation of ASCII code) is cross-over’d with Key (Binary representation) generated by BBS (Section 3.1) Here, The Cipher text that will be sent consists of 2 numbers A’D and C’B instead of AB in the reverse number (if single crossover is performed) and DB’F and AE’C instead of ABC in the reverse order (if double crossover is performed). For the plain text “TEXT” the encryption process is as follows:

Character ASCII Value Binary Value T 83 01010011 E 69 01000101 X 87 01010111 T 83 01010011

01010011(83) 01011001 01001001(73) è(Crossover) è(Mutation) 00011001(25) 00010011 00000011(3) So the Cipher Text is (3, 73). This process is continued until all the text in source file (Plain text) is completed.

3.2.5 Decryption The Decryption process is carried out as : Generate random numbers concurrently. for every bit in the file(cipher text) until EOF if random number generated is odd read two characters at a time.

perform CROSSOVER between the second number read and the xi(binary representations) where the breaking point is n%k . perform MUTATION at the (2*xi)%k position in the crossovered numbers.


57

perform CROSSOVER between first offspring of the above phase and the first character read(in binary representations).

else if the number generated is even perform Double CROSSOVER between the second number and the key (binary representation of ASCII value)where the first breaking point is xi%k and second one is (xi+s)%k . perform MUTATION at the (2*xi)%k position in the crossovered numbers. perform CROSSOVER between first number of the above output and the first character read(binary representations).

The first number of the above output is the plain text(if single crossover is to be performed)

A’ D A’ D A D è(Crossover) è(Mutation)

C D C’ D C’ D A D A B à Plain Text

è(Crossover) C’ B C’ D If double crossover is to be performed

D B’ F D B’ F D B F

è(Double Crossover) è(Mutation) D E F D E F D E F

D B F A B C à Plain Text è(Double Crossover) A E’ C D E’ F

4. Analysis The proposed algorithm has the following advantages : Suitable for hardware or software: Algorithm uses only primitive computational operations that can be easily implemented on hardware in a less economic way. Which is not possible with RSA and Triple DES. Variable-length key : The key length can be varied in the algorithm which is possible in RSA but not in Triple DES. Low memory Requirement : A low memory requirement makes the proposed algorithm suitable for smart cards and other devices with restricted memory which is not possible in RSA and Triple-DES. Resistant to Known Plain Text, Known Cipher Text and Bruteforce Attacks Resistant to Timing Attack : As the algorithm uses Blum, Blum, Shub generator for Key Generation it is resistant to timing attacks(Section 2.1). RSA is prone to this kind of attack but Triple-DES is not. Computationally Secure : As the proposed algorithm maps each character in plain text to two characters in cipher text. It is hard to break the cipher. This feature is present in both the RSA and Triple-DES algorithms.

Ease of analysis : The algorithm is explained concisely over here. Even though it is difficult to cryptanalyze . RSA and DES lacks in this feature.

5. Conclusion and Future Enhancements Hence, The paper proposes a new algorithm which is

equivalently secure with RSA and Triple DES and which can be easily implemented on the hardware.

Future process will be devoted to extend the algorithm to achieve the other security services like Authentication, Data Integrity etc.,

References [1] Lenore Blum, Manuel Blum, and Michael Shub.,

“Comparision of two pseudo random number generators” Proc. CRYPTO’82, pages. 61-78, Newyork, 1983.

[2] William Stallings, “Cryptography and Network Security”, Prentice Hall, 3rd Edition.

[3] Subramil Som, Jyotsna Kumar Mandal and Soumya Basu, “A Genetic Functions Based Cryptosystem (GFC)”, IJCSNS, September 2009.

[4] Ankit Fadia, “Network Security”, Macmillan India Ltd.

Authors Profile

Praneeth Kumar G received the B.Tech Degree in Computer Science and Engineering from Progressive Engineering College in 2008. During May’ 2008 – Aug’ 2009, he worked in Concepts in Computing(CIC) as a Software Engineer. He is presently working at C V S R College of Engineering as an Assistant Professor. His areas of interest include software engineering and Information Security.

Vishnu Murthy G received the B.Tech. and M.Tech. degrees in Computer Science and Engineering. He is resource person for IEG and Birla Off campus programmes. He is presently pursuing his Ph.D in J.N.T.U. and heading the Department of Computer Science and Engineering in C V S R College Of Engineering. His areas of interest include software Engineering, Information Security and Image Processing.


58

Cluster Management using Cluster Size Ratio in Ad Hoc Networks

D K L V Chandra Mouly, Ch D V Subba Rao and M M Naidu

Department of Computer Science and Engineering

S V University College of Engineering, Tirupati - 517502, India. [email protected], [email protected]

Abstract: Cluster Management using Cluster Size Ratio (CMCSR) is a completely distributed algorithm for partitioning a given set of mobile nodes into clusters. The proposed algorithm tries to reduce the amount of computational and information overhead while maintaining a stable cluster formation. It constructs and maintains a backbone topology based on a minimal dominating set (MDS) of the network. According to this algorithm, each node determines the membership in the MDS for itself and its one-hop neighbors based on one-hop neighbor information that is disseminated among neighboring nodes using willingness and priority information of the nodes. The algorithm then ensures that the members of the MDS are connected into a connected dominating set (CDS), which can be used to form the backbone infrastructure of the communication network to facilitate routing. The algorithm outperforms the existing algorithms with respect to stability. Load balancing the cluster heads using the cluster size ratio is the heuristic used in this algorithm.

1. Introduction This section discusses elementary issues of ad hoc networks and benefits of clustering.

1.1 Ad Hoc Networks In the next generation of wireless communication systems, there will be a need for the rapid deployment of independent mobile users. Significant examples include establishing survivable, efficient, dynamic communication for emergency/ rescue operations, disaster relief efforts, and military networks. Such network scenarios cannot rely on centralized and organized connectivity, and can be conceived as applications of ad hoc networks. An ad hoc network is an autonomous collection of mobile users that communicate over relatively bandwidth constrained wireless links. Since the nodes are mobile, the network topology may change rapidly and unpredictably over time. The network is decentralized, where all network activity including discovering the topology and delivering messages will be taken care by the nodes, i.e., routing functionality will be incorporated into mobile nodes. The set of applications for ad hoc networks is diverse, ranging from small, static networks that are constrained by power sources, to large-scale, mobile, highly dynamic networks. The design of network protocols for these networks is a complex issue. Regardless of the application, ad hoc networks need efficient distributed algorithms to determine network organization, link scheduling, and routing. However, determining viable routing paths and

delivering messages in a decentralized environment where network topology fluctuates is not a welldefined problem [1].

1.2 Clustering in Ad Hoc Networks A wireless ad hoc network consists of nodes that move freely and communicate with each other using wireless links. Ad-hoc networks do not use specialized routers for path discovery and traffic routing. One way to support efficient communication between nodes is to develop wireless backbone architecture; this means that certain nodes must be selected to form the backbone. Over time, the backbone must change to reflect the changes in the network topology as nodes move around. The algorithm that selects the members of the backbone should naturally be fast, but also should require as little communication between nodes as possible, since mobile nodes are often powered by batteries. One way to solve this problem is to group the nodes into clusters, where one node in each cluster functions as cluster head, responsible for routing [2].

1.3 Benefits of clustering Ad-hoc networks are suited for use in situations where an infrastructure is unavailable or to deploy one is not cost effective. One of many possible uses of mobile ad-hoc networks is in some business environments, where the need for collaborative computing might be more important outside the office environment than inside, such as in business meeting outside the office to brief clients on a given assignment. Mobile ad-hoc networks allow the construction of flexible and adaptive networks with no fixed infrastructure. These networks are expected to play an important role in the future wireless generation. Future wireless technology will require highly-adaptive mobile networking technology to effectively manage multi-hop ad-hoc network clusters, which will not only operate autonomously but also will be able to attach at some point to the fixed networks. 2. Literature Review This section emphasizes some of the past clustering techniques.

2.1 Types of Topology Management There are two approaches to topology management in ad hoc networks:

• Power control. • Hierarchical topology organization.


59

2.1.1 Power Control Different Power control mechanisms adjust the power on a per-node basis, so that one-hop neighbor connectivity is balanced and overall network connectivity is ensured [4, 5, 6]. Li [7] proved that network connectivity is minimally maintained as long as the decreased power level keeps at least one neighbor remaining connected at every 2π/3 to 5π/6 angular separation. Ramanathan [7] proposed to incrementally adjust nodes power levels so as to keep network connectivity at each step topologies derived from power-control schemes often result in unidirectional links that create harmful interference due to the different transmission ranges among one-hop neighbors [9]. The dependencies on volatile information in mobile networks, such as node locations [4], signal strength or angular positions [8] also contribute to the instability of topology control algorithms based on power control.

2.1.2 Hierarchical topology control This approach to topology control is often called clustering, and consists of selecting a set of cluster heads in a way that every node is associated with a cluster head, and cluster heads are connected with one another directly or by means of gateways, so that the union of gateways and cluster heads constitute a connected backbone [10, 14, 15]. Once elected, the cluster heads and the gateways helps to reduce the complexity of maintaining topology information, and can simplify such essential functions as routing, bandwidth allocation, channel access, power control or virtual circuit support. For clustering to be effective, the links and nodes that are part of the backbone (i.e., cluster heads, gateways, and the links that connect them) must be close to minimum and must also be connected.

2.2 TYPES OF CLUSTER HEAD ELECTIONS Cluster heads can be elected in four ways.

• Deterministic Clustering • Non-Deterministic Clustering • Reactive Clustering • Proactive Clustering

A. Deterministic Clustering Deterministic clustering can determine the cluster heads in a single round. Different heuristics have been used to form clusters and to elect cluster heads. Several approaches [12] utilized the node identifiers to elect the cluster heads within one or multiple hops.

B. Non-Deterministic Clustering In non-deterministic clustering, negotiations are used. Negotiations require multiple incremental steps, and may incur an election jitter during the process, because of the lack of consensus about the nodes being elected as the cluster heads. Examples of this approach are the “core” extraction algorithm [13] and the spanning tree algorithm [14]. SPAN [13] allows a node to delay the announcement of becoming a cluster head for random amounts of time to attempt to attain minimum conflicts between cluster heads in its one-hop neighborhood.

C. Reactive Clustering It is on-demand clustering algorithm. There is no periodic exchange of clustering information in the network. Instead, whenever there is data traffic, cluster related information is piggybacked in outgoing data packets and extracted out of received packets.

D. Proactive Clustering Some are proactive clustering algorithms, which require periodic broadcast of cluster-related information. SPAN [13] adaptively elects coordinators according to the remaining energy and the number of pairs of neighbors a node can connect.

2.3 Scope for Present Work

The efficiency of a communication network depends not only on its control protocols, but also on its topology. Our work i.e. CMCSR proposes a distributed topology management algorithm that constructs and maintains a backbone topology based on a Minimal Dominating Set (MDS) of the network. Without topology management each and every node should maintain the routing information for all the nodes they need. By using topology management a subset of nodes are selected called cluster heads and each cluster head performs the routing work for its members. 3. System Model

3.1 Assumptions This work assumes that an ad hoc network comprises a group of mobile nodes communicating through a common broadcast channel using omni-directional antennas with the same transmission range. The topology of an ad hoc network is thus presented by an undirected graph G = (V,E), where V is the set of network nodes, and E ⊆ U * V is the set of links between nodes. The existence of a link (u, v) ∈ E also means (v; u) ∈ E, and that nodes u and v are within the packet-reception range of each other, in which case u and v are called one-hop neighbors of each other. The set of one- hop neighbors of a node i is denoted by Ni. Two nodes that are not connected but share at least one common one-hop neighbor are called two-hop neighbor of each other. Each node has one unique identifier, and all transmissions are omni directional with the same transmission range. The nodes move with constant mobility. The energy is decreased linearly. Different types of nodes consume energy at different rates. We ignore the energy consumed due to local computations, but assume that the energy consumption rate is only dependent on the type of the node. A host consumes 0.6% of the total energy per minute in these algorithms; whereas a cluster head consumes 3%.

3.2 Model

In an ad hoc network all nodes are alike and all are mobile. There are no base stations to coordinate the activities of subsets of nodes. Therefore, all the nodes have to collectively make decisions. All communication is over wireless links. A wireless link can be established between a pair of nodes only if they are within wireless range of each


60

other. We will only consider bidirectional links. It is assumed the MAC layer will mask unidirectional links and pass only bidirectional links. Beacons could be used to determine the presence of neighboring nodes. After the absence of some number of successive beacons from a neighboring node, it is concluded that the node is no longer a neighbor. Two nodes that have a wireless link will, henceforth, be said to be one wireless hop away from each other. They are also said to be immediate neighbors. Communication between nodes is over a single shared channel. In ad hoc networks the nodes within each neighborhood are not known a priori. The individual cluster may transition to spatial TDMA for inter-cluster and intra-cluster communication. All nodes broadcast their node identity periodically to maintain neighborhood integrity. Due to mobility, a node’s neighborhood changes with time. As the mobility of nodes may not be predictable, changes in network topology over time are arbitrary. However, nodes may not be aware of changes in their neighborhood. Therefore, clusters and cluster heads must be updated frequently to maintain accurate network topology.

3.2.1 Attributes of a node

The attributes of a node and their functionality are as given in Table I.

Table 1: Attributes of a node and their functionality

ATTRIBUTE FUNCTION

ID Unique name given to node ENERGY The capacity to work in MOBILITY The speed of the node when it

is moving WILLINGNESS How much the node is willing

to be a cluster head PRIORITY Has the priority among other

nodes to became a cluster head CLUSTER SIZE Cluster size ratio which it is

having TYPE Whether it is cluster head or

gateway or door way or member

NEIGHBORS Number of one – hop neighbors

3.2.2 Computing Priorities of Nodes

Given that cluster heads provide the backbone for a number of network control functions, their energy consumption is more pronounced than that of ordinary hosts. Low-energy nodes must try to avoid serving as cluster heads to save energy. However, to balance the load of serving as cluster heads, every node should take the responsibility of serving as a cluster head for some period of time with some likelihood. Furthermore, node mobility has to be considered in cluster head elections. To take into account the mobility and energy levels of nodes in their election, we define the two-hop neighbor information needed to assign node

priorities that consists of two components: (a) Neighboring Nodes, (b) Willingness value assigned to a node as a function of its mobility and energy level. We denote the willingness value of node i by Wi, the speed of node i by a scalar Mi that ranges from 0 to 1 meters per second, and the remaining energy on node i as Ei in the range of 0 and 1. The willingness Wi is a function that should be defined according to the following criteria: 1. To enhance survivability, each node should have the responsibility of serving as a cluster head with some nonzero probability determined by its willingness value. 2. To facilitate with the stability and the frequency with which cluster head elections must take place, the willingness value of a node should remain constant as long as the variation of the speed and energy level of the node do not exceed some threshold values. 3. To avoid electing cluster heads that quickly lose connectivity with their neighbors after being elected, the willingness value of a node should decrease drastically after the mobility of the node exceeds a given value. 4. To prolong the battery life of a node, its willingness value should decrease drastically after the remaining energy of the node drops below the given level. Willingness value (Wi) is as specified below:

Wi = 2log2(Ei+.9)log2(Mi+2) Here the constants 0.9 and 2 in Eq. (1) eliminate the boundary conditions in the logarithmic operations. The logarithmic operations on the speed and the remaining energy values render higher willingness values in the high energy and low speed field, while giving close to zero values in the low energy and high-speed region. Priority value (Pi) is a function of no.of neighbors and willingness i.prio = 2log2(Wi)/n Figure 1 illustrates the effect of the two factors on the priority values. From the Figure 2 we can conclude that the priority is directly proportional to the willingness value and number of neighbors.

00.10.20.30.40.50.60.70.80.9

1

Priority

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.91

4

7

10

Willingness

Number of Neighbors

Figure 1. Priority Graph


61

3.2.3 MDS and Cluster Head Election

The approach to establishing a minimal dominating set (MDS) is based on three key observations First, using negotiations among nodes to establish which nodes should begin the MDS incurs substantial overhead when nodes move around and the quality of links changes frequently. Hence, nodes should be allowed to make MDS membership decisions based on local information. Second, because in an MDS every node is one hop away from a cluster head, the local information needed at any node needs to include only nodes that are one and two hops away from the node itself. Third, having too many cluster heads around the same set of nodes does not lead to an MDS. Hence, to attain a selection of nodes to the MDS without negotiation, nodes should rank one another using the two-hop neighborhood information they need. Based on the above, the approach adopted in CMCSR consists of each node communicating to its neighbor’s information about all its two-hop neighbors. Using this information, each node computes a priority for each node in its two-hop neighborhood, such that no two nodes can have the same priority at the same instant of time. A node can become cluster head if the node has highest priority in its two hop neighborhood.

3.2.4 Connected Dominating Set Election

The CDS [4] of a network topology is constructed in two steps. In the first step, if two cluster heads in the MDS are separated by three hops and there are no other cluster heads between them, a node with the highest priority on the shortest paths between the two cluster heads is elected as a doorway, and is added to the CDS. Therefore, the addition of a doorway brings the connected components in which the two cluster heads reside one hop closer. In the second step, if two cluster heads or one cluster head and one doorway node are only two hops away and there are no other cluster heads between them, one of the nodes between them with the highest priority becomes a gateway to connect cluster head to cluster head or doorway to cluster head. After these steps, the CDS is formed. CDS is constructed in two steps

• Selecting doorway • Selecting gateway

(a) Selecting Doorway Node i can become a doorway for cluster heads n and j if the following conditions are satisfied. i) If cluster n and j are not two hops away. ii) There is no other cluster head m on the shortest path between n and j. iii) There is no other node m with higher priority than node i. Selecting gateway

Node i can become a gateway for cluster head n and j if the following conditions are satisfied. i) If there is no cluster head or doorway between n and j. ii) If there is no node with higher priority than node i.

3.2.5 Computing Cluster Size ratio

The objective is to develop an enhancement for existing heuristics to provide a contiguous balance of loading on the elected cluster heads. Once a node is elected as cluster head it is desirable for it to stay as a cluster head up to some maximum specified amount of time, or budget. The budget is a user defined constraint placed on the heuristic and can be modified to meet the unique characteristics of the system, i.e., the battery life of individual nodes. Some of the goals of the heuristic are: 1. Minimize the number and size of the data structures required to implement the heuristic, 2. Extend the cluster head duration budget based on an input parameter, 3. Allow every node equal opportunity to become a cluster head in time, 4. Maximize the stability in the network. Data Structures The data structures necessary for the heuristic consist of one local variable: Physical ID (PID). The PID is the initial id given and is unique for each individual node. However, this changes with time to represent the elect ability of a node. Basic Idea The node id load heuristic operates on the principle of load balancing. That is, the ids of each non-cluster head node cycles through the queue at a rate of 1 unit per run of the load-balancing heuristic. Each node has a minimum value of 0 and a maximum value of Max_Cluster Size. Upon reaching Max_Cluster Size a node will rotate to a value of 0 on the next cluster election heuristic run. As the cluster election heuristics run they will use the priorities to determine the cluster heads of the network. A cluster head will maintain this value until it has exhausted its cluster head duration budget. At this point it will set its work to 0, i.e., less than any other node, and become a normal node. 4. Performance Evaluation We have conducted simulation experiments to evaluate the performance of the proposed heuristic i.e. CMCSR. These simulation results were then compared against Topology Management by Priority Ordering TMPO [15]. We assumed a variety of systems running with 10, 20, 40, 60, 80 and 100 nodes to simulate ad hoc networks with varying levels of node density. Two nodes are said to have a wireless link between them if they are within communication range of each other. Additionally, the span of a cluster, i.e. the maximum number of wireless hops between a node and its cluster head (d) was set to 2. The entire simulation was conducted in a 1150 * 1150 unit region. Initially, each node was assigned a unique node id and (x, y) coordinates within the region. The nodes were then allowed to move at random in any direction at a speed of not greater than half of the wireless range of a node per second. The simulation range is set to 2000 seconds, and the network was sampled for every 2 seconds. At each sample time the proposed cluster size ratio and cluster election heuristic was run to determine cluster heads and their associated clusters. Each simulation run for 2000 seconds measures several performance metrics. The main simulation metric measured was Cluster head


62

Duration, which provides a basis for evaluating the performance of the proposed load-balancing heuristic. For the purposes of these simulations we have set the cluster head budget to be a function of the maximum amount of work it performs That is, once a node becomes a cluster head it will remain a cluster head until it has exhausted its maximum work load, or until it loses out to another cluster head based on the rules of the cluster election heuristic. The proposed CMCSR algorithm makes a noticeable difference in the cluster head duration (ranging from 4% to 28%). This shows that the load-balancing heuristics generates longer cluster head durations; it will also produce much tighter and more deterministic responses (stability). These results are not surprising. Therefore, once a cluster head is elected it continues as cluster head for a maximum of the programmed budget. This will provide the longer cluster head durations that we see. The cluster size ratio heuristic is continuously rotating, moving ordinary nodes into the position of becoming a cluster head. Therefore, once a cluster head budget is exceeded, a different cluster head is elected and the process repeats. This provides the cluster size ratio effect of distributing the responsibility of being a cluster head among all nodes. We present below three graphs for our simulation results. First one is the average cluster head duration. Second one is the average number of cluster head. And finally the improvement graph for the cluster head duration.

4.1 Nodes Vs Cluster Head Duration

Figure 2 shows the graph for the average cluster head duration. X-axis takes the number of nodes and y-axis shows the cluster head duration in seconds. The topology management is executed for 1800 seconds for each x nodes and the values are noted. Totally the program is executed for 18000 seconds. The diamond shaped line indicates the cluster head duration without load i.e. incase of TMPO.

0

5

10

15

20

25

30

10 20 30 40 50 60 70 80 90 100

No. of Nodes

Clu

ster

Hea

d D

urat

ion(

in S

ec)

TMPO CMCSR

Figure 2. Average Cluster head duration Vs no. of nodes

Second the topology management is executed for the 600 seconds for each x nodes and the values are noted. Totally the program is executed for 18000 seconds. The square

shaped line indicates the cluster head duration with load i..e. incase of CMCSR.

4.2 Nodes Vs Number of Cluster Heads

Figure 3 shows the graph for the average number of cluster heads formed during the topology management. The topology management is executed for 1800 seconds for each of x nodes and the values are noted. The diamond shaped line indicates the number of cluster head formed during topology management without load (TMPO). Second the topology management executed for the 600 seconds for each x nodes and the values are noted. Totally the program is executed for 18000 seconds. The square shaped line indicates the cluster head formed during topology management with load (CMCSR).

0

2

4

6

8

10

12

14

16

10 20 30 40 50 60 70 80 90 100

No.of Nodes

No.

of C

lust

ers

TMPO CMCSR

Figure 3. Average no. of clusters

Figure 4. Average Cluster head duration

0

5

10

15

20

25

900 1800 2700 3600System Executed(Sec)

TMPO CMCSR


63

4.3 Improvement in Cluster head Duration

Figure 4 shows the graph for the average cluster head duration. X-axis takes system executed in seconds and y-axis shows the average cluster head duration in seconds. The below graph is constructed under the following conditions. Both TMPO and CMCSR is run for 900 sec, 1800 sec, 2700 sec and 3600 sec by taking total number of nodes into account as 50. The diamond shaped line indicates the cluster head duration without load and the square shaped line indicates the cluster head duration with load. The results related to the above three graphs indicates that CMCSR outperforms TMPO.

5. Conclusions The cluster size load balancing heuristics have been proposed for ad hoc networks. The cluster election heuristics favor the election of cluster heads based on node willingness and number of neighbors. Here the heuristic places a cluster size budget on the contiguous amount of time that a node acts as cluster head. As seen from the simulation results, this heuristic produce larger cluster head durations while decreasing the cluster size and enhancing the stability. Our proposed CMCSR is a novel energy-aware topology management approach based on dynamic node priorities and cluster size load in ad hoc networks. CMCSR consists of two parts that implement the MDS and CDS elections respectively. Compared to five prior heuristics of MDS and CDS elections in ad hoc networks, MDS offers four key advantages. i) CMCSR obtains the MDS and CDS of the network without any negotiation stage; only two-hop neighbor information is needed. ii) CMCSR allows nodes in the network to periodically re-compute their priorities, so as to balance the cluster head role and prolong the battery life of each node. iii) CMCSR introduces the willingness value of a node, which decides the probability of the node being elected into the MDS according to the battery life and mobility of the node and iv) MDS introduces doorway concept for the CDS in addition to the well-known gateway and cluster head concepts. A key contribution of this work consists of converting the static attributes of a node, such as node identifier, into a dynamic control mechanism that incorporates the three key factors for topology management in ad hoc networks -- the nodal battery life, mobility, and cluster size load balancing. Although existing proposals have addressed all these aspects, CMCSR constitutes a more comprehensive approach. References [1] http://w3.antd.nist.gov/wahn_mahn.shtml. [2] Tomas Johansson and Lenka Carr-Motyˇckov´. “On

Clustering in Ad Hoc Networks,” First Swedish National Computer Networking Workshop, SNCNW2003, 8-10 September, 2003.

[3] R.Pandian, P.Seethalakshmi and V.Ramachandran, “Enhanced Routing Protocol for Video Transmission over Mobile Adhoc Network,” Journal of Applied

Sciences Research 2(6): 336-340, INSInet Publication, 2006.

[4] L. Hu. “Topology Control for Multihop Packet Radio Networks,”. IEEE Transactions on Communications, 41(10), Oct. 1993.

[5] S. Narayanaswamy, V. Kawadia, R. S. Sreenivas, and P. R. Kumar, “Power Control in Ad-Hoc Networks: Theory, Architecture, Algorithm and Implementation of the COMPOW Protocol,” Proceedings of the European Wireless Conference on Next Generation Wireless Networks: Technologies, Protocols, Services and Applications, pages 156-162, Florence, Italy, Feb. 25-28, 2002.

[6] H. Takagi and L. Kleinrock, “Optimal Transmission Ranges for Randomly Distributed Packet Radio Terminals,” IEEE Transactions on Communications, 32(3),7, Mar. 1984.

[7] L. Li, V. Bahl, Y.M. Wang, and R. Wattenhofer, “Distributed Topology Control for Power Efficient Operation in Multihop Wireless Ad Hoc Networks,” Proceedings of IEEE Conference on Computer Communications (INFOCOM), Apr. 2001.

[8] R. Ramanathan and R. Rosales-Hain, “Topology Control of Multihop Wireless Networks using Transmit Power Adjustment,” Proceedings of IEEE Conference on Computer Communications (INFOCOM), IEEE, Mar. 26-30, 2000.

[9] R. Prakash, “Unidirectional Links Prove Costly in Wireless Ad-Hoc Networks,” Proceedings of the Discrete Algorithms and Methods for Mobile Computing and Communications - DialM, Seattle, WA, Aug. 20, 1999.

[10] S. Bandyopadhyay and E. J. Coyle, “An Energy Efficient Hierarchical Clustering Algorithm for Wireless Sensor Networks”, In Proc. INFOCOM 2003, San Francisco, Apr, 2003.

[11] M. Maeda and Ed Callaway, "Cluster Tree Protocol (ver.0.6)",http://www.ieee802.org/15/pub/2001/May01/01189r0P80215_ TG4-Cluster-Tree-Network.pdf.

[12] L. Bao and J.J. Garcia-Luna-Aceves, “Transmission Scheduling in Ad Hoc Networks with Directional Antennas,” Proc. ACM Eighth Annual International Conference on Mobile Computing and networking, Atlanta, Georgia, USA, Sep, 23-28 2002.

[13] B. Chen, K. Jamieson, H. Balakrishnan, and R. Morris, “Span: an Energy-Efficient Coordination Algorithm for Topology Maintenance in Ad Hoc Wireless Networks,” In Proc. 7th ACM MOBICOM, Rome, Italy, Jul, 2001.

[14] C.C. Chiang, H.K. Wu, W. Liu, and M. Gerla, “Routing in Clustered Multihop, Mobile Wireless Networks with Fading Channel,” IEEE Singapore International Conference on Networks SICON'97, pages 197-211, Singapore, Apr. 14-17, 1997.

[15] L. Bao and J.J. Garcia-Luna-Aceves, “Topology Management in Ad Hoc Networks,” Proc of the 4th ACM Interational Symposium on Mobile Ad Hoc Networking and Computing (MOBIHOC), Annapolis, Maryland, USA, Jun. 2003.


64

Authors Profile Mr D K L V Chandra Mouly received M.Tech (CSE) from S V University College of Engineering, Tirupati, India in the year 2007. Currently he is pursuing his Ph.D. (Part-time) at S V University, Tirupati. His areas of interests are Computer Networks and Distributed Systems. Dr Ch D V Subba Rao received Ph.D (CSE) from S V University, Tirupati, India in 2008. He got 18 years of teaching experience. At present, he is working as Associate Professor, Dept of Computer Science and Engineering, S V University College of Engineering, Tirupati, India. His areas of interests are Distributed Systems, Operating Systems, Computer Networks and Programming Language Concepts. Dr M M Naidu received Ph.D (IIT-Delhi) in the year 1988. He got 32 years of teaching experience. Currently he is working as Professor in the Dept of Computer Science and Engineering, S V University College of Engineering, Tirupati, India. His areas of interests include Software Engineering, Enterprise Resource Planning, Computer Networks and Computer Graphics.


65

Analysis and Proposal of a Geocast Routing Algorithm Intended For Street Lighting System

Based On Wireless Sensor Networks Rodrigo Palucci Pantoni1 and Dennis Brandão1

1Eng. School of São Carlos of University of São Paulo,

São Carlos, Brazil [email protected] and [email protected]

Abstract: This work is part of research project where are studied and developed efficient technologies highlighted by the ReLuz (Brazilian National Program of Public Lightning and Efficient Light Signalization) program. It is proposed, in that context, the development of remote command and monitoring infrastructure for management purposes for large areas. Specifically, it refers the research and analysis of efficient routing algorithms in terms of energy consumption, guarantee of package delivery and good performance (minimum delay). Two new geocast routing algorithms are implemented and compared with other candidates applied to street lighting system. Such algorithms refer to wireless sensor networks IEEE 802.15.4 through multi-hop and low reach communication, and low cost. The results show that the proposed algorithm GGPSRII saves more energy, presents a good percentage of delivery guarantee, good performance than compared ones.

Keywords: Lighting system, Geocast routing algorithm, IEEE 802.15.4, Wireless sensor network.

1. Introduction Improvements in the quality of public lighting systems have a direct impact in the quality of life for a large part of the population as well as in the efficiency and rationality of using electric power. This work is part of the research project that intends to integrate the study and development of efficient technologies in the scope of the Brazilian federal government program ReLuz (National Program of Efficient Public Lighting and Traffic Lights) [1]: a telecommand system for managing public lighting in large areas. The expected result is the economic operation of a public lighting system with an economy index superior to the indexes currently registered in the mentioned program, due to the efficiency provided by the use of high performance electronic systems together with a telecommand system. The telecommand system is composed by devices attached to the points of light, which are interconnected via network, and software tools used for monitoring and control. This work refers to the research and analysis for choosing and implementing efficient network routing algorithms in terms of electric power consumption, that is, besides the rational use of the electric power, there is also the concern about saving energy in terms of network communication. The routing algorithm refers to IEEE 802.15.4 mesh-based wireless sensor networks (WSN) through multi-hop communication based on low range communication, low cost and minimum electric power consumption, with

distributed routing for retransmitting information to the final destination. The IEEE 802.15.4 standard was chosen because of its minimum consumption, low cost and protocol simplicity. The main challenge for this work is handling the limitations of the IEEE 802.15.4 with the characteristics of dense networks, trying to reach a balance among guarantee of delivery, performance (minimum delay) and electric power efficiency for the specific purpose of public lighting.

2. Correlated works of public lighting system Several proposals found in the literature related to public lighting use PLC technology (Power Line Communication) [2, 3, 4]. However, there are a few limitations from this technology when it is applied in public lighting [3, 5], such as noise and impedance variations. Furthermore, Brazilian infrastructure would have to be in perfect state to achieve a good operation, but in reality the infrastructure is old and would have to be rebuilt. A company [6] applied the physical and data link layer standard of the model ISO/OSI IEEE 802.11 to the wireless network protocol, which has higher electric power consumption and high communication rates. In this solution, each point of light has a network point that communicates and sends information to an Internet point. In [7], it was proposed the use of ZigBee [8] as the protocol for the lighting system. [9] mentioned that the use of ZigBee is not suitable for public lighting systems, although it has not been quantitatively proved. They suggested the use of the network protocol 6LoWPAN [10] for devices IEEE 802.15.4 [11] and a GPS. 6LoWPAN is destined to devices with low cost, low electric power waste and low communication rates, and the main goal is to integrate the Internet with WSN naturally. However, the use of 6LoWPAN over the layers defined in IEEE 802.15.4 does not define a routing algorithm by itself. Nevertheless, such proposal naturally matches the work hereby proposed.

3. Proposed public lighting system To create the routing algorithm, it is necessary to visualize the entire architecture, in brief, where the requirements are displayed superficially. The requirements are: points supervision (device status, whether it is connected to the network or not; battery power; life time estimation for


66

battery and lamp; and LEDs luminosity level), control (switch the luminosity level lamp on or off; switch on/off a lamp post, a selected segment, a street, neighbourhood, city, etc; automatic programmed actions and freedom to actuate the devices through a remote tool) and diagnostic and alarms (trigger an event when a network, hardware or lamp failure occurs). During each device’s initialization, a GPS set attributes the geographic coordinates of their positions. The user would be able to select a specific area on the map besides the previously programmed area (for example, selecting a segment from a street) to actuate (switch the lights on or off). Figure 1 shows the area selected by the user through the system supervision and control tool.

Figure 1. System Supervision and Control Screen

Mechanisms to automatize input information process in the public lighting system must be applied, in order to become the process simplest and fast. Such mechanisms are not in the scope of this paper.

4. Correlated works of routing algorithms for WSN

This section intends to review researches found in the literature related to routing algorithms for WSN in general, independent of the application. Routing algorithms for mesh-based networks, such as AODV (Ad-Hoc On Demand Distance Vector) [12], DSR (Dynamic Source Routing) [13], and DREAM (Distance Routing Effect Algorithm for Mobility) [13] were developed to provide mobility. These algorithms are reactive, that is, routes are determined by flooding through nodes searching the addressee node when a flow of information (triggered by the upper layer) occurs. When the route is determined, it is stored in the memory of the participant nodes. This mechanism causes high energy costs, performance costs and guarantee of delivery costs. Besides, devices would have to keep large routing tables in dense networks, which would be impossible considering that such devices have low memory available. It is also interesting to keep the minimum possible overhead in the network package, because it is limited at 127 Kb by the IEEE 802.15.4 specification on the data link

layer level. In contrast to reactive algorithms, there are proactive algorithms, such as DSDV (Dynamic Destination Sequenced Distance-Vector) [14] and OLSR (Optimized Link State Routing) [15]. Instead of building optimized routes only when they are necessary, proactive algorithms keep a matrix of connections always updated, which reflects the current network status and channels available for data transfer. From the computational and electric power consumption point of view, these algorithms are too costly, especially when providing mobility or when a fail occurs. Geographic routing algorithms use the geographic location of the devices as a reference, and the location can be obtained from a GPS. The great advantage for this type of routing is that routing tables are not necessary because the devices decide where to forward the package according to the smallest Euclidean distance of the destination coordinate, for example. Since those algorithms were designed for mobile devices, one of the steps for this type of routing is transmitting “hello” messages to all neighbouring devices (in radio range), that periodically send packages with the identification (such as the network address) of the device and its position. So, the devices store the location of their neighbours. They apply the greedy routing [16] by transmitting a message to their neighbours that are relatively closer in distance to the final destination. For any variation of the greedy algorithm, it is important to define a discard criterion to prevent the message from being transmitted uninterruptedly over the network in case the specified destination is not located. However, in cases where it is necessary to find a balance between performance and guarantee of delivery, the discard criterion must be defined even if there is a path to the final addressee. To assure package delivery, greedy algorithms are frequently used combined with recovering strategies, providing two operation modes. Such strategies are used when a package is discarded in “pure” greedy mode, in case there is an obstacle or a non-operating network device, for example. The most prominent recovering strategy uses planar graphs. Basically, the idea is to draw the network as a unique graph on a plane and forward the message in the direction of the adjacent faces, which consequently forward the package to the final destination. Those strategies are extensively studied, as in GFG (Greedy-Face-Greedy) [17], GPVFR (Greedy Path Vector Face Routing) [18], GPSR (Greedy Perimeter Stateless Routing) [19] and GOAFR++ (Greedy Other Adaptive Face Routing plus plus) [20]. In the GPSR algorithm, the recovering strategy is named perimeter mode and uses the right hand rule to direct the flow of network packages through the devices. In case the distance from the device to the destination is smaller than the distance to its neighbours, the algorithm returns to the greedy mode. The term unicast means a point-to-point connection where data is sent from a sender to a receiver. The most


67

appropriate type of routing in our context, however, is the so-called geocast, that also requires the devices to know their geographic positions via GPS. The algorithms deliver network messages to the devices in a specific geographic area, delivering a message from one device to many devices. There are several routing algorithms developed for that purpose, some based on flooding messages, directed flooding messages, and without flooding. Flooding messages algorithms find the path to the destination area the same way AODV does. The first package arriving at the destination area broadcasts to all nodes in the area. On the other hand, directed flooding messages algorithms define two types of areas, the destination and the routing area. An example of directed flooding is the LBM (Location Based Multicast) algorithm [21] that is executed as follows: the routing area is defined as an area in the direction of the destination area, and packages forwarded outside these two areas are discarded.

5. Proposed Routing Algorithms The algorithms proposed in this study, without flooding, are named GGPSR (Geocast Greedy Perimeter Stateless Routing Protocol) and GGPSR II (Geocast Greedy Perimeter Stateless Routing Protocol II). They consist of two parts: modified GPSR to find the destination area and geocast to broadcast the message to all addressee devices. Instead of using the specific coordinate of a device as the destination, the central point of the destination area is calculated. The package is then forwarded to this point and, when it gets to the destination area, the first device receiving the message broadcasts it to all devices in the area. As soon as a device receives the broadcasted message, the device checks if it has already received this message, checking a sequential number and therefore avoiding unnecessary retransmissions. In case the device has not received the message yet but it is in the destination area, the device receives and rebroadcasts the message to the network. Table I shows the pseudo-algorithm GGPSR in a simplified way. Considering the GPSR part of the proposed algorithm GGPSR (only), this work suggests some modifications for lighting systems applications. The first modification is related to the “hello” messages. For the discussed application, devices are fixed. Initially, it was assumed that only one “hello” message when the device is initialized would be enough. However, it is interesting to keep this functionality but with a periodicity much longer than what it is used in mobile devices. The reason to still keep this periodicity is that the device can simply stop operating because of a permanent or temporary failure caused by an obstruction. Information about the neighbors from the devices can affect the network reliability because each device is also a message router. The second modification is related to storing geographic positions for the neighbors (applied to GGPSR and GGPSRII). Originally, GPSR has three types of messages: “hello” messages, consultation to destination locations and data messages [19]. “Hello” messages are responsible to inform the new device’s location to its neighbors; data

messages are responsible to forward the packages; and, finally, consultation messages are responsible to obtain the location of the addressees from one or more location databases for a certain unique device’s identification (for example, the network address). Once the scenario does not have mobility, consultation messages and location servers can be removed. Such functionality was implemented through “hello” messages that send the unique identification and the location. Besides, the supervision and control software requires the location of all devices on the network. Thus, the package would be sent by the system with the geographic coordinate of the destination, instead of having only the network address and requiring the current device to obtain its location through the location server. The difference between GGPSR and GGPSRII consists only in the trigger condition of “hello” messages. In GGPSR, the trigger is invocated according to a pre-determinated frequency (period). On the other hand, in GGPSRII, the trigger is invocated if only the data message does not reach the destination (geocast region).Thus, it is necessary to implement a confirmation message to inform the data message forwarding failure. Table II show the GGPSRII simplified pseudo-algorithm.

Table 1: GGPSR simplified pseudo-algorithm // Initialization For all devices Send_Broadcast_Hello_Neighbors (); Start Hello_Timer (period); If Hello_Timer_Expire Send_Broadcast_Hello_Neighbors (); //Send If (Packet.Destination_Position != myPosition && myPosition = = UNICAST) ModifiedGPSR_Forward (Packet); Else If (Packet.Destination_Position = = myPosition && myPosition = = GEOCAST){ If (Packet.seqN_ < ReceivedSeqNo){ Broadcast_Neighborhood_Geocast_Region(Packet); } } //Receive If (Packet.Destination_Position == myPosition && myPosition = = UNICAST) ModifiedGPSR_Receive (Packet); Else If (Packet.Destination_Position = = myPosition && myPosition = = GEOCAST){ If (Packet.seqNo¬_ < ReceivedSeqNo){ ModifiedGPSR_Receive (Packet); }

} Regarding the destination area, it can have the shape of a four-vertex polygon, circle and point (in this last case, the communication is unicast). The geographic position is represented through geodesic coordinates (latitude and longitude). Each coordinate is allocated as floating types, which in language C has 4 bytes and precision of seven decimal places. In relation to the value ranges, this size is more than enough: the field "hours" of the latitude coordinate varies between -180 and 180, whereas longitude varies between -90 and 90. Table III shows the header struct in C language of the packet types, including “hello” messages (hdr_gpsr_hello), and data messages (hdr_gpsr_data) of the proposed protocol.


68

Table 2: GGPSRII simplified pseudo-algorithm // Initialization For all devices Send_Broadcast_Hello_Neighbors (); //Send If (Packet.Source_Position = = myPosition && myPosition == UNICAST){ ModifiedGPSR_Forward (Packet); Start Confirmation_Timer (period); } Else If (Packet.Destination_Position != myPosition && myPosition == UNICAST) ModifiedGPSR_Forward (Packet); Else If (Packet.Destination_Position == myPosition && myPosition == GEOCAST){ If (Packet.seqNo_ < ReceivedSeqNo){ Broadcast_Neighborhood_Geocast_Region(Packet); } } //Receive If (Packet.Destination_Position = = myPosition && myPosition == UNICAST) ModifiedGPSR_Receive (Packet); Else If (Packet.Destination_Position == myPosition && myPosition == GEOCAST){ If (Packet.seqNo_ < ReceivedSeqNo){ ModifiedGPSR_Receive (Packet); } } //Any Time If Confirmation _Timer_Expire Send_Broadcast_Hello_Neighbors ();

Related to the “hello” packet, the field “type_” refers to the packet type (whether it is “hello” or data packet). The field “x_” and “y_” are the source device geodesic coordinates. The field “seqNo” is used to control the receiving and rebroadcast actions, i. e., a device must have to broadcast just a once “hello” messages to its neighbours. Related to the data packet, the fields “sx_” and “sy_” are the source device geodesic coordinates. The field “ts_” is the timestamp, used for calculate latency. The fields “sx_GF_Failed” and “sy_GF_Failed_” correspond to the coordinates where the greedy mode have failed, that are used for the packet whether it can be return to that mode in case it is in the perimeter mode. The field “seqNo” is used for control the receiving and rebroadcast actions on the geocast region. The rest of variables are the polygon coordinates’.

Table 3: GGPSR and GGPSRII header packets struct hdr_gpsr_hello { u_int8_t type_; float x_; float y_; int seqNo_; }; struct hdr_gpsr_data { u_int8_t type_; u_int8_t mode_; float sx_; float sy_; float ts_; float sx_GF_Failed_; float sy_GF_Failed_; float dst_x1; float dst_y1; float dst_x2; float dst_y2; float dst_x3; float dst_y3; float dst_x4; float dst_y4; int seqNo_; };

6. Simulation and results The simulation was analyzed to obtain quantitative data in order to decide which routing algorithm would be more appropriate to the problem in discussion. First, the ZigBee protocol (which uses AODV routing algorithm) was evaluated because it is a consolidated standard that provides interoperability among several manufacturers and reduced costs for production. For this reason, the AODV protocol is simulated and compared to the GPSR, GGSR, GGPSRII and LBM protocols. Simulations were performed on a largely used simulator for the academic area, the ns-2.33 [22]. Originally, in relation to routing algorithms, ns-2.33 only implemented the AODV algorithm. Hence, the authors implemented GPSR, LBM, GGPSR and GGPSRII protocols. Configurations for simulating all routing protocols are presented in Table IV. Devices are equally distributed vertically and horizontally along 50 meters to simulate lamp posts in a simplified way. Figure 2 shows the location of the devices. Data always flows from device 98 to all nodes on the “last line” (devices 0, 1, 2, 3, 4, 5, 6, 7, 8 and 9). So, device 98 is the network coordinator. In addition, the ns2 oTcl code was implemented to keep the energy of device 98 always as its initial energy.

Table 4: Configurations for simulation Network Interface Phy/WirelessPhy/802_15_4 MAC Mac/802_15_4 IFQ Queue/DropTail/PriQueue Link Layer LL Antena Antenna/OmniAntenna Dimension X 170 Dimension Y 270 IFQLEN 50 Propagation Propagation/TwoRayGround Phy/WirelessPhy Pt_ 7.214e-3 (100m) Number of Devices 100 Duration 1000 simulation time Transmission Power 0.3 mW Reception Power 0.28 mW Initial Energy 1 Joule Packet size (less header)

64 bytes

Flow CBR (Constant Bit Rate) Simulation foresees the basic operation situation: network traffic is requested every twelve hours (assuming that the unit of simulation is in hours), that is, switch the lights on and off for a street segment, for example. GPSR and GGPSR “Hello” messages periodicity were configured to 12 hours, that is the time to send data. It is important to emphasize that in case of unicast algorithms, ten messages are sent from device 98 to all devices in the “last line”, as mentioned before. In case of geocast algorithms, only one message is sent to all addressees. Figure 3 shows a comparison of the electric power of all devices in the network summed during the time interval. It was verified that the AODV protocol is the least efficient, and GGPSRII is the most efficient. It can be concluded that the use of ZigBee protocol is strongly not recommended.


69

Figure 2. Simulated scenario

Figure 4 shows the rate in percentage of throughput to each evaluated protocol. For the AODV traffic, it was necessary to configure the beginning of the information flow in a non-simultaneous way, with a short period of time between transmissions. If all messages were sent to the ten devices simultaneously, no message would be successfully received. The scheduling configuration was also applied to the GPSR routing for comparison, even though GPSR does not need this artifice. Observe that because of the AODV flooding characteristic, there is a very significant lost of messages that reassures discredit on the ZigBee protocol. For this same reason, it can be verified in Figure 5 that there is a delay average much higher than the AODV algorithm.

Figure 3. Comparing the network total energy

Figure 4. Delivery rate

Figure 5. Delay average (performance)

Table V shows the packet header comparison. GGPSRII presents lesser size of bytes then “LBM – request”, which is the largest header in terms of bytes.

TABLE 5: Packet header Comparison

Packet Header Size (bytes) AODV - request 28 AODV - response 26 AODV - error 14 AODV - confirmation 4 LBM - request 62 LBM - response 52 LBM - error 14 LBM - confirmation 4 GPSR – hello 12 GPSR - data 34 GGPSRII - hello 12 GGPSRII - data 58

7. Conclusions and future works The proposed GGPSRII routing algorithm proved to be efficient in terms of energy, and provides a good balance in terms of performance and guarantee of delivery. Quantitative results showed that the research has a solid base, that is, resources can be invested in hardware prototypes for devices (an alternative to ZigBee) and in implementing the supervision and control software. Later on, this algorithm will be implemented to transmit a message to multiple areas, using the Fermat point concept proposed by [23]. Final validation tests for the protocol will be simulated in real scenarios obtained from lamp posts mapping, usually archived in city halls. The authors will also develop an alarm and network diagnostic mechanism for the lighting system based on the protocol of the proposed routing. Further on, application layer services will be included to switch the lights, trigger alarms, provide diagnostics, etc.

Acknowledgment The authors gratefully acknowledge the academic support and research structure from the Engineering School of São Carlos - University of São Paulo. The authors also acknowledge the important technical contributions from Smar International Corporation and the Prof. Tracy Camp for helping in provide an implementation of LBM algorithm, which was very helpful for an implementation of the LBM implementation of this work.


70

References [1] Reluz Program (Jan. 2010). “National Program of

Efficient Public Lighting and Traffic Lights”. Available: http://www.eletrobras.gov.br/ EM_Programas_Reluz

[2] Ei-Shirbeeny, E.H.T.; Bakka, M.E.. “Experimental pilot project for automating street lighting system in Abu Dhabi using powerline communication”. Proceedings of the 10th IEEE International Conference on Electronics, Circuits and Systems. Vol. 2, p.743 – 746, Dec. 2003

[3] Chueiri, I.J.; Bianchim, C.G. “Sistema de Comando e Controle de Potência em Grupo para Iluminação Pública”. BR n.PI0201334-7, 2002.

[4] Sungkwan C.; Dhingra, V. “Street lighting control based on LonWorks power line communication Power Line Communications and Its Applications”. IEEE Symposium, 2008.

[5] Sutterlin, P.; Downey, W. (1999). “A power line communication tutorial – challenges and technologies”, Technical Report, Echelon Corporation, 1999.

[6] Streetlight Intelligence (Jan. 2010). Available: http://www.streetlightiq.com

[7] Barriquello, C.H. ; Garcia, J.M. ; Corrêa, C. ; Menezes, C.V. ; Campos, A. ; Do Prado, R.N. “Sistema Inteligente Baseado em Zigbee para Iluminação Pública com Lâmpadas de LEDS”. In: XVII Congresso Brasileiro de Automática. Anais do XVII CBA. Juiz de Fora, 2008.

[8] Zigbee. ZigBee PRO Specification, ZigBee Alliance. 2007.

[9] Denardin, G.W.; Barriquello, C.H.; Campos, A. Do Prado, R.N. “An Intelligent System for Street Lighting Monitoring and Control”. 10° Congresso Brasileiro de Eletrônica de Potência. Brasil, Bonito, 2009.

[10] Kushalnagar, N.; Montenegro, G.; Schumacher, C.. “IPv6 over Low-Power Wireless Personal Area Networks (6LoWPANs): Overview, Assumptions, Problem Statement, and Goals”. Request for Comments: 4919, 2007.

[11] IEEE 802.15.4. “Wireless Medium Access Control (MAC) and Physical Layer (PHY) Specifications for Low-Rate Wireless Personal Area Networks”, IEEE Computer Society, 2006.

[12] Perkins, C. E.; Belding-Royer, E. M.; Das, S. R.. “Ad Hoc On-Demand Distance Vector Routing”, Request for Comments: 3561, 2003.

[13] Basagni, S.; Chlamtac, I.; Syrotiuk, V.R.; Woodward, B.A. “A distance routing effect algorithm for mobility (dream)”. In Proceedings of ACM/IEEE MobiCom ’98, 1998.

[14] Perkins, C.; Bhagwat, P. “Highly Dynamic Destination Sequenced Distance-Vector Routing for Mobile Computers”. Comp. Commun. 1994.

[15] Clausen, T.; Jacquet, P. “Optimized Link State Routing Protocol”. Request for Comments: 3626, 2003.

[16] Finn, G.G. “Routing and addressing problems in large metropolitan scale internetworks”. Technical Report ISI/RR-87-180, ISI, 1987.

[17] Bose, P.; Morin, P., Stojmenovic, I.; Urrutia, J.”Routing with guaranteed delivery in ad hoc wireless networks”. In: Proceedings of the 3rd International Workshop on Discrete algorithms and methods for mobile computing and communications. 1999, ACM Press.

[18] Leong, B.; Mitra, S.; Liskov. B. “Path vector face routing: Geographic routing with local face information”. In Proceedings of the IEEE Conference on Network Protocols, 2005.

[19] Karp, B; Kung, H. T. (2000) GPSR: greedy perimeter stateless routing for wireless networks. in Proceedings of the 6th ACM/IEEE MobiCom. 2000, pp. 243-254, ACM Press.

[20] Kuhn,F.;Wattenhofer,R.;Zhang,Y.;Zollinger, A. “Geometric ad-hoc routing: of theory and practice”. in Proceedings of the 22nd annual symposium on principles of distributed computing, 2003.

[21] Ko, Y.; Vaidya, N.H.”Geocasting in mobile ahoc networks: Location-based multicast algnrithms” .In Proceedings of WMCSA, pages 101-110, 1999.

[22] Network Simulator NS2 (Jan. 2010). Available: http://www.isi.edu/nsnam/ns

[23] Lee,S.;Ko,Y. “Geometry-driven Scheme for Geocast Routing in Mobile Ad Hoc Networks”. The 2006 IEEE 63rd Vehicular Technology Conference (VTC), Melbourne, Australia, 2006.

Authors Profile Rodrigo Palucci Pantoni R&D Systems Analyst, received the Computer Science degree in 2000 and subsequently received the M.S. in 2006 at the University of São Paulo (USP). He's attending the Ph.D course, in the same university, as part of his job at the Smar R&D department in the area

of software development for automation control and fieldbuses. He joined Smar in 2000, working in the Smar R&D department where he conducts research and development of host systems, including a Fieldbus Foundation Asset Management and a Configurator system. He now teaches computer networks at Information Systems course at University Dr. Francisco Maeda. Dennis Brandão He received his Ph.D. degree in mechanical

engineering at the University of São Paulo in 2005. He now teaches “Industrial Automation” at the Department of Electrical Engineering of the same university. His research activities are mainly in the area of fieldbus technology and application, with a particular interest for distributed systems and continuous process control.


71

Routing protocols in wireless networks

1Vineet Agrawal , 2Dr Yashpal Singh,3Manish Varshney ,4Vidushi Gupta 1Reader Deptt of Computer Science & Engg RBCET Bareilly, India

Email [email protected] 2Reader Deptt of Computer Science & Engg BIET Jhansi, India

Email [email protected] 3Sr Lecturer Deptt of Computer Science & Engg, SRMSWCET Bareilly, India

Email [email protected] 4Lecturer Deptt of Computer Science & Engg, SRMSWCET Bareilly, India

Email [email protected] Abstract: An ad hoc mobile network is a group of mobile nodes that are dynamically and arbitrarily located in such a manner that the interconnections between nodes are capable of changing on a repetitive basis. In order to facilitate communication within the network, a routing protocol is used to discover routes between nodes. The primary goal of such an ad hoc network routing protocol is correct and efficient route establishment between a pair of nodes so that messages may be delivered in a timely manner. In this paper we examine routing protocols for ad hoc networks and evaluate these protocols based on a given set of parameters. The scope of paper was to test routing performance of four different routing protocols (AODV, DSR, DSDV,etc) in variable network sizes up to thousand nodes. Various types of scenarios are generated and each of the protocol is simulated on each of these, then their parameters like throughput, packet delivery ratio and delay will be compared The performance differentials are analyzed using varying pause time, constant nodes and dynamic topology. Based on the observations, we make valuable conclusions about which protocol performs better in which condition. Keywords: Ad hoc ,Table-Driven, Demand Driven AODV, DSDV,DSR, , MANET.

1. Introduction

Mobile hosts and wireless networking hardware are becoming widely available, and extensive work has been done recently in integrating these elements into conventional networks such as the Internet. Oftentimes, however, mobile users will want to communicate in situations in which no fixed wired infrastructure such as this is available, either because it may not be economically practical or physically possible to provide the necessary infrastructure or because the expediency of the situation does not permit its installation. In networks comprised entirely of wireless stations, communication between source and destination nodes may require traversal of multiple hops, as radio ranges are finite. Since Routing Protocols emergence in the 1970s, wireless networks have become increasingly popular in the computing industry. This is particularly true within the past decade, which has seen wireless networks being adapted to enable mobility. There are currently two variations of mobile wireless networks. The first is known as the infrastructure network (i.e., a network with fixed and wired gateways).

Figure 1. A simple ad hoc network of three wireless mobile

hosts The bridges for these networks are known as base stations. A mobile unit within these networks connects to, and communicates with, the nearest base station that is within its communication radius. As the mobile travels out of range of one base station and into the range of another, a “handoff” occurs from the old base station to the new, and the mobile is able to continue communication seamlessly throughout the network. Typical applications of this type of network include office wireless local area networks (WLANs). The second type of mobile wireless network is the infrastructure less mobile network, commonly known as an ad hoc network. Infrastructure less networks has no fixed routers; all nodes are capable of movement and can be connected dynamically in an arbitrary manner. Nodes of these networks function as routers which discover and maintain routes to other nodes in the network. Example applications of ad hoc networks are emergency search-and-rescue operations, meetings or conventions in which persons wish to quickly share information, and data acquisition operations in inhospitable terrain. A community of ad hoc network researchers has proposed, implemented, and measured a variety of routing algorithms for such networks. The observation that topology changes more rapidly on a mobile, wireless network than on wired networks, where the use of Distance Vector (DV), Link State (LS), and Path Vector routing algorithms is well established, motivates this body of work . DV and LS algorithms require continual distribution of a current map of the entire network’s topology to all routers. DV’s Bellman- Ford approach constructs this global picture transitively; each router includes its distance from all network destinations in each of its periodic beacons. LS’s Dijkstra approach directly floods announcements of the change in any link’s status to every router in the network. Small inaccuracies in the state at a router under both DV and LS can cause routing loops or disconnection [7,9]. When the topology is in constant flux, as under mobility, LS generates torrents of link status


72

change messages, and DV either suffers from out-of-date state [4], or generates torrents of triggered updates. The two dominant factors in the scaling of a routing algorithm are: The rate of change of the topology and The number of routers in the routing domain. Both factors affect the message complexity of DV and LS routing algorithms: intuitively, pushing current state globally costs packets proportional to the product of the rate of state change and number of destinations for the updated state. Hierarchy is the most widely deployed approach to scale routing as the number of network destinations increases. Without hierarchy, Internet routing could not scale to support today’s number of Internet leaf networks. An Autonomous System runs an intra-domain routing protocol inside its borders, and appears as a single entity in the backbone inter-domain routing protocol, BGP. This hierarchy is based on well-defined and rarely changing administrative and topological boundaries. It is therefore not easily applicable to freely moving ad-hoc wireless networks, where topology has no well-defined AS boundaries, and routers may have no common administrative authority. Caching has come to prominence as a strategy for scaling ad-hoc routing protocols. Instead, routers running these protocols request topological information in an on-demand fashion as required by their packet forwarding load, and cache it aggressively. When their cached topological information becomes out-of date, these routers must obtain more current topological information to continue routing successfully. Because of the fact that it may be necessary to hop several hops (multi-hop) before a packet reaches the destination, a routing protocol is needed. The routing protocol has two main functions, selection of routes for various source-destination pairs and the delivery of messages to their correct destination. The second function is conceptually straightforward using a variety of protocols and data structures (routing tables). This paper is focused on selecting and finding routes. This paper examines routing protocols designed for these wireless networks by first describing the operation of each of the protocols and then comparing their various characteristics. The remainder of the paper is organized as follows. The next section presents a discussion of two subdivisions of ad hoc routing protocols. Another section discusses current table-driven protocols, while a later section describes those protocols which are classified as on-demand. The paper then presents Simulation parameters and the performance evaluation including, a general comparison of table-driven and on-demand protocols. Finally, the last section concludes the paper.

2. Ad Hoc Routing Protocols Since the advent of Defense Advanced Research Projects Agency (DARPA) packet radio networks in the early 1970s [1], several protocols have been developed for ad hoc mobile networks. Such protocols must deal with the typical limitations of these networks, which include high power consumption, low bandwidth, and high error rates. As shown in Fig.2, these routing protocols may generally be categorized as: • Table-driven(proactive) • Source-initiated (demand-driven)(reactive)

2.1 Table-Driven Routing Protocols Table-driven routing protocols attempt to maintain reliable, up-to-date routing information from each node to every other node in the network. These protocols require each node to maintain one or more tables to store routing information, and they react to changes in network topology by propagating updates throughout the network in order to maintain a consistent network view. The areas in which they differ are the number of necessary routing-related tables and the methods by which changes in network structure are broadcast. The following sections discuss some of the existing table-driven ad hoc routing protocols. 2.1.1 Destination-Sequenced Distance-Vector Routing The Destination-Sequenced Distance-Vector Routing protocol (DSDV) described in [2] is a table-driven algorithm based on the classical Bellman-Ford routing mechanism [3]. The developments made to the Bellman-Ford algorithm include freedom from loops in routing tables.

Figure 2.Categorization of ad hoc routing protocols

DSDV [2] is a hop-by-hop distance vector routing protocol that in each node has a routing table that for all available destinations stores the next-hop and number of hops for that destination Every mobile node in the network maintains a routing table in which all of the possible destinations within the network and the number of hops to each destination are recorded. Each entry is marked with a sequence number assigned by the destination node. The sequence numbers enable the mobile nodes to distinguish stale routes from new ones, thereby avoiding the formation of routing loops. Routing table updates are periodically transmitted throughout the network in order to maintain table consistency. DSDV basically is distance vector with small adjustments to make it better suited for ad-hoc networks. These modifications consist of triggered updates that will take care of topology changes in the time between broadcasts. To reduce the amount of information in these packets there are two types of update messages defined: full and incremental dump. The full dump carries all available routing information and the incremental dump that only carries the information that has changed since the last dump. Because DSDV is dependent on periodic broadcasts it needs some time to converge before a route can be used. This converge time can probably be considered negligible in a static wired network, where the topology is not changing so frequently. In an ad-hoc network on the other hand, where the topology is expected to be very dynamic, this converge time will probably mean a lot of dropped packets before a valid route is detected. The periodic broadcasts also add a large amount of overhead into the network. New route broadcasts contain the address of the destination, the


73

number of hops to reach the destination, the sequence number of the information received regarding the destination, as well as a new sequence number unique to the broadcast [2].. 2.1.2The Wireless Routing Protocol The Wireless Routing Protocol (WRP) described in [5] is a table-based protocol with the goal of maintaining routing information among all nodes in the network. To describe WRP, we model a network as an undirected graph represented as G.V; E. , where V is the set of nodes and E is the set of links (or edges) connecting the nodes. Each node represents a router and is a computing unit involving a processor, local memory and input and output queues with unlimited capacity. In a wireless network, a node has radio connectivity with multiple nodes and a single physical radio link connects a node with many other nodes. Each node in the network is responsible for maintaining four tables: • Distance table • Routing table • Link-cost table • Message retransmission list (MRL) table Each entry of the MRL of the update Message, a retransmission counter, an acknowledgment- required flag vector with one entry per neighbor, and a list of updates sent in the update message. A link is assumed to exist between two nodes only if there is radio connectivity between the two nodes and they can exchange update messages reliably with a certain probability of success The MRL records which updates in an update message need to be retransmitted and which neighbors should acknowledge the retransmission [5]. Mobiles inform each other of link changes through the use of update messages. An update message is sent only between neighboring nodes and contains a list of updates (the destination, the distance to the destination, and the predecessor of the destination), as well as a list of responses indicating which mobiles should acknowledge (ACK) the update. Mobiles send update messages after processing updates from neighbors or detecting a change in a link to a neighbor. In the event of the loss of a link between two nodes, the nodes send update messages to their neighbors. The neighbors then modify their distance table entries and check for new possible paths through other nodes. Nodes learn of the existence of their neighbors from the receipt of acknowledgments and other messages. If a node is not sending messages, it must send a hello message within a specified time period to ensure connectivity. Otherwise, the lack of messages from the node indicates the failure of that link; this may cause a false alarm. Because of the broadcast nature of the radio channel, a node can send a single update message to inform all its neighbors about changes in its routing table; however, each such neighbor sends an ACK to the originator node. To ensure that connectivity with a neighbor still exists when there are no recent transmissions of routing table updates or ACKs, periodic update messages without any routing table changes (null update messages) are sent to the neighbors. The time interval between two such null update messages is the HelloInterval. If a node fails to receive any type of message from a neighbor for a specified amount of time (e.g., three or four times the HelloInterval known as the Router Dead-Interval), the node

must assume that connectivity with that neighbor has been lost. When a mobile receives a hello message from a new node, that new node is added to the mobile’s routing table, and the mobile sends the new node a copy of its routing table information. Part of the novelty of WRP stems from the way in which it achieves loop freedom. In WRP, routing nodes communicate the distance and second-to-last hop information for each destination in the wireless networks. WRP belongs to the class of path-finding algorithms with an important exception. It avoids the “count-to-infinity” problem [6] by forcing each node to perform consistency checks of predecessor information reported by all its neighbors. This ultimately (although not instantaneously) eliminates looping situations and provides faster route convergence when a link failure event occurs. 2.2 Source-Initiated On-Demand Routing A different approach from table-driven routing is source-initiated on-demand routing. This type of routing creates routes only when desired by the source node. When a node requires a route to a destination, it initiates a route discovery process within the network. This process is completed once a route is found or all possible route permutations have been examined. Once a route has been established, it is maintained by a route maintenance procedure until either the destination becomes inaccessible along every path from the source or until the route is no longer desired. 2.2.1 Ad Hoc On-Demand Distance Vector Routing (AODV) The Ad Hoc On-Demand Distance Vector routing protocol (AODV) is an improvement of the Destination-Sequenced Distance Vector routing protocol (DSDV)1. DSDV has its efficiency in creating smaller ad-hoc networks. Since it requires periodic advertisement and global dissemination of connectivity information for correct operation, it leads to recurrent system-wide broadcasts. Therefore the size of DSDV ad-hoc networks is strongly limited. When using DSDV, every mobile node also needs to maintain a whole list of routes for each destination within the mobile network. The advantage of AODV is that it tries to reduce the number of required broadcasts. It creates the routes on an on-demand basis, as opposed to maintain a complete list of routes for each destination. Therefore, the authors of AODV classify it as a pure on-demand route acquisition system [3]. 2.2.1.1 Path Discovery Process When trying to send a message to a destination node without knowing an active route2 to it, the sending node will initiate a path discovery process. A route request message (RREQ) is broadcasted to all neighbors, which persist to broadcast the message to their neighbors and so on. The forwarding process is continued until the destination node is reached or until an intermediate node knows a route to the destination that is new enough. To ensure loop-free and most recent route information, every node maintains two counters: sequence number and broadcast_id. The broadcast_id and the address of the source node uniquely identify a RREQ message. broadcast_id is incremented for every RREQ the source


74

node initiates. An intermediate node can receive multiple copies of the same route request broadcast from various neighbors. In this case – if a node has already received a RREQ with the same source address and broadcast_id – it will discard the packet without broadcasting it furthermore. When an intermediate node forwards the RREQ message, it records the address of the neighbor from which it received the first copy of the broadcast packet. This way, the reverse path from all nodes back to the source is being built automatically. The RREQ packet contains two sequence numbers: the source sequence number and the last destination sequence number known to the source. The source sequence number is used to maintain “freshness” information about the reverse route to the source while the destination sequence number specifies what actuality a route to the destination must have before it is accepted by the source. [3] When the route request broadcast reaches the destination or an intermediate node with a fresh enough route, the node responds by sending a unicast route reply packet (RREP) back to the node from which it received the RREQ. So actually the packet is sent back reverse the path built during broadcast forwarding. A route is considered fresh enough, if the intermediate node’s route to the destination node has a destination sequence number which is equal or greater than the one contained in the RREQ packet. As the RREP is sent back to the source, every intermediate node along this path adds a forward route entry to its routing table. The forward route is set active for some time indicated by a route timer entry. The default value is 3000 milliseconds, as referred in the AODV RFC [4]. If the route is no longer used, it will be deleted after the specified amount of time. Since the RREP packet is always sent back the reverse path established by the routing request, AODV only supports symmetric links.

Figure.3. AODV Path Discovery Process.

2.2.2 Dynamic Source Routing The Dynamic Source Routing (DSR) protocol presented in [8] is an on-demand routing protocol that is based on the concept of source routing. The Dynamic Source Routing (DSR) protocol is an on-demand routing protocol based on source routing. In the source routing technique, a sender determines the correct sequence of nodes through Dynamic Source Routing (DSR) [3][12][13] also belongs to the class of reactive protocols and allows nodes to dynamically

discover a route across multiple network hops to any destination. DSR uses no periodic routing messages, thereby reducing network bandwidth overhead, conserving battery power and avoiding large routing updates throughout the ad-hoc network. Instead DSR relies on support from the MAC layer (the MAC layer should inform the routing protocol about link failures). The two basic modes of operation in DSR are route discovery and route maintenance.

Figure 4. AODV Route Maintenance by using Link

failure Notification Message

2.2.2.1Route Discovery Route discovery allows any host in the ad hoc network to dynamically find out a route to any other host in the ad hoc network, whether directly reachable within wireless transmission range or reachable through one or more intermediate network hops through other hosts. A host initiating a route discovery broadcasts a route request packet which may be received by those hosts within wireless transmission range of it. The route request packet identifies the host, refer red to as the target of the route discovery, for which the route is requested. If the route discovery is successful the initiating host receives a route reply packet listing a sequence of network hops through which it may reach the target. In addition to the address of the original initiator of the request and the target of the request, each route request packet contains a route record, in which is accumulated a record of the sequence of hops taken by the route request packet as it is propagated through the ad hoc network during this route discovery. Each route request packet also contains a unique request id, set by the initiator from a locally-maintained sequence number. In order to detect each duplicate route requests received, host in the ad hoc network maintains a list of the h initiator address, request id i pairs that it has recently received on any route request. 2.2.2.2 Route Maintenance Route maintenance can be accomplished by two different processes: • Hop-by-hop acknowledgement at the data link layer • End-to-end acknowledgements Hop-by-hop acknowledgement at the data link layer allows an early detection and retransmission of lost or corrupt packets. If the data link layer determines a fatal transmission error (for example, because the maximum number of retransmissions is exceeded), a route error packet is being sent back to the sender of the packet. The route error packet contains two parts of information: The address of the node detecting the error and the host’s address which it was trying to transmit the packet to. Whenever a node


75

receives a route error packet, the hop in error is removed from the route cache and all routes containing this hop are truncated at that point. End-to-end acknowledgement may be used, if wireless transmission between two hosts does not work equally well in both directions. As long as a route exists by which the two end hosts are able to communicate, route maintenance is possible. There may be different routes in both directions. In this case, replies or acknowledgements on the application or transport layer may be used to indicate the status of the route from one host to the other. However, with end-to-end acknowledgement it is not possible to find out the hop which has been in error. 3 .Simulation And Its Parameters 3.1 Methodology The main concentration of the project was to test the ability of different routing protocols to respond on network topology changes (for instance link breaks, node movement, and so on). Furthermore the focus was set on different network sizes, varying number of nodes and area sizes. Our investigations did not include the protocol’s operation under heavy load, e.g. its operation in congestion situations. Therefore only rather small packet sizes and one source node were selected. As referenced in many other papers, Our protocol evaluations are based on the simulation of 50 wireless nodes forming an ad hoc network, moving about over a rectangular (1500m X 300m) flat space for 200 seconds of simulated time. We chose a rectangular space in order to force the use of longer routes between nodes than would occur in a square space with equal node density. In order to enable direct, fair comparisons between the protocols, it was critical to challenge the protocols with identical loads and environmental conditions. Each run of the simulator accepts as input a scenario file that describes the exact motion of each node and the exact sequence of packets originated by each node, together with the exact time at which each change in motion or packet origination is to occur. We pre-generated 9 different scenario files with varying movement patterns and traffic loads, and then ran all three routing protocols against each of these scenario files. Since each protocol was challenged in an identical fashion, we can directly compare the performance results of the protocols 3.2 Mobility Model An important factor in mobile ad-hoc networks is the movement of nodes, which is characterized by speed, direction and rate of change. Mobility in the “physical world” is unpredictable, often unrepeatable, and it has a dramatic effect on the protocols developed to support node movement. Therefore, different “synthetic” types of mobility models have been proposed to simulate new protocols. Synthetic means to realistically represent node movement, but without using network traces. Nodes in the simulation move according to a model that we call the “random waypoint” model. The movement scenario files we used for each simulation are characterized by a pause time. Each node begins the simulation by remaining stationary for pause time seconds. It then selects a random destination in the 1500m x 300m space and moves to that destination at a

speed distributed uniformly between 0 and some maximum speed. Upon reaching the destination, the node pauses again for pause time seconds, selects another destination, and proceeds there as previously described, repeating this behavior for the duration of the simulation. Each simulation ran for 200 seconds of simulated time. We ran our simulations with movement patterns generated for 9 different pause times: 2, 10, 15, 25, 35, 50, 75, 85, 100 seconds. A pause time of 0 seconds corresponds to continuous motion, and a pause time of 200 (the length of the simulation) corresponds to no motion. Hence reducing pause time increases mobility. In this way we put our protocols in networks of varying mobility. Because the performance of the protocols is very sensitive to movement pattern, we generated scenario files with 9 different pause times. All routing protocols were run on the same 9 scenario files. We report in this paper data from simulations using a maximum node speed of 20 meters per second (average speed 10 meters per second). 3.3 Communication Model As the purpose of our simulation was to compare the performance of each routing protocol, we select our traffic sources to be constant bit rate (CBR) sources. When defining the parameters of the communication model, we experimented with sending rates of 3 packets per second, networks containing maximum connection of 35, and packet sizes of 512 bytes. All communication patterns were peer-to-peer, and connections were started at times uniformly distributed between 0 and 180 seconds. The 9 different scenario files for maximum node movement speed (20 m/s) moving in a random waypoint model with which we compared the routing protocols. 3.4 Performance Metrics In order to compare routing protocols, the following performance metrics are considered: • Throughput: a dimensional parameter which gives the portion of the channel capacity used for useful transmission selects a destination at the beginning of the simulation and (i.e., data packets correctly delivered to the destinations). • Average End to End delay: the average end-to-end delay of data packets, i.e. the period between the data packet generation time and the time when the last bit arrives at the destination. •Packet delivery ratio: the ratio among the number of packets received by the TCP descends at the final destination and the number of packets originated by the “application layer” sources. It is a measure of efficiency of the protocol 4. Performances Analysis DSDV which is a table driven proactive routing protocol completely prevails over the on demand reactive routing protocols AODV and DSR .Since DSDV proactively maintains the routes to all destination in its table it does not have to initiate the route request process as frequently as in AODV and DSR while sending packets. Hence on average DSDV clearly has less delay. Now we can easily examine


76

that DSR is the worst protocol in terms of delay. At high mobility and more network load (512 byte packets at 3 packets/sec) insistent route caching strategy of DSR fails. In these stressful condition links break very often leading to invalidation of routes cached .Hence in these conditions, picking up of staled cached routes occur leading to utilization of additional network bandwidth and interface queue slots even though the packet is ultimately dropped, leading to more delay DSR performed inefficiently in our metrics (PDR and Throughput) in these “stressful” situations (higher mobility, more network load). The reason of these phenomena is the aggressive use of route caching in DSR. In our observation, such caching provides a significant benefit up to a certain extent. With higher loads the degree of caching is deemed too large to benefit performance. Often, stale routes are chosen since route length (and not any freshness criterion) is the only metric used to pick routes from the cache when faced with multiple choices. Picking stale routes causes two problems: • Consumption of additional network bandwidth and interface queue slots even though the packet is eventually dropped or delayed • Possible pollution of caches in other nodes With high mobility, the possibilities of the caches being stale are quite high in DSR. Eventually when a route discovery is initiated, the large number of replies (as all RREQs are replied) received in response is associated with higher MAC overhead and cause increased interference to data traffic. Hence, the cache staleness and high MAC overhead mutually result in significant degradation in performance for DSR in high mobility An efficient mechanism to remove stale cached routes can improve performance of DSR. On other hand since in AODV only the first arriving request packet (RREQs) is answered and further no RREQs are answered therefore it leads to less no. of replies (RREPs) .Also the error packets RERRs are are broadcasted in AODV which leads to lesser MAC load as compared to unicasted REERs of DSR which leads to much MAC layer load.

Figure 4. Average End To End Delay

5. Comparisons The subsequent sections provide comparisons of the previously described routing algorithms. The next section compares table-driven protocols, and a further section compares on demand protocols. 5.1 Table-Driven vs. On-Demand Routing As discussed former, the table-driven ad hoc routing border on is similar to the connectionless approach of forwarding

packets, with no regard to when and how frequently such routes are preferred. It relies on an underlying routing table revise mechanism that involves the stable propagation of routing information. This is not the case, however, for on-demand routing protocols. When a node using an

Figure 5. Throughput of Receiving Packets

Figure 6. Packet Delivery Ratios

desires a route to a new destination, it will have to wait until such a route can be discovered. On the other hand, since routing information is constantly propagated and maintained in table-driven routing protocols, a route to every other node in the ad hoc network is always available, regardless of whether or not it is needed. This feature, although useful for datagram traffic, incurs substantial signaling traffic and power consumption. Since both bandwidth and battery power are scarce resources in mobile computers, this becomes a serious limitation. 6. Conclusion In this paper we provide descriptions of several routing schemes proposed for ad hoc mobile networks. We also provide a classification of these schemes according to the routing strategy (i.e., table-driven and on-demand). We have presented a comparison of these two categories of routing protocols, highlighting their features, differences, and characteristics .We has compared the performance of DSDV, AODV and DSR We used a detailed simulation model to demonstrate the performance characteristics of these protocols. By simulating we can argue that if delay is our main criteria than DSDV can be our best choice But if reliability and throughput are our main parameters for selection then AODV gives better results compare to others because its throughput and packet delivery ratio is best among others. While there are many other issues that need to be considered in analyzing the performance of ad hoc networks, we believe that our work could provide intuition for future protocol selection and analysis in ad hoc


77

networks. While we focus only on the network throughput, reliability and the delay, it would be interesting to consider other metrics like power consumption, the number of hops to route the packet, fault tolerance, minimizing the number of control packets etc. Parameters On-demand Table-den References

[1] J. Jubin and J. Tornow, “The DARPA Packet Radio Network Protocols,” Proc. IEEE, vol. 75, no. 1, 1987, pp. 21–32.

[2] C. E. Perkins and P. Bhagwat, “Highly Dynamic Destination-Sequenced Distance-Vector Routing (DSDV) for Mobile Computers,” Comp. Commun.Rev., Oct. 1994, pp. 234–44.

[3] L. R. Ford Jr. and D. R. Fulkerson, Flows in Networks, Princeton Univ. Press, 1962.

[4] C. Perkins, E. Belding-Royer, and S. Das, “RFC 3561: Ad hoc on-demand distance vector (AODV) routing,” July 2003, category: experimental. [Online]. Available: ftp://ftp.isi.edu/in-notes/rfc3561.txt

[5] S. Murthy and J. J. Garcia-Luna-Aceves, “An Efficient Routing Protocol for Wireless Networks,” ACM Mobile Networks and App. J., Special Issue on Routing in Mobile Communication Networks, Oct. 1996, pp. 183–97.

[6] A. S. Tanenbaum, Computer Networks, 3rd ed., Ch. 5, Englewood Cliffs, NJ: Prentice Hall, 1996, pp. 357–58.

[7] C. E. Perkins and E. M. Royer, “Ad-hoc On-Demand Distance Vector Routing,” Proc. 2nd IEEE Wksp. Mobile Comp. Sys. and Apps., Feb. 1999, pp. 90–100.

[8] D. B. Johnson and D. A. Maltz, “Dynamic Source Routing in Ad-HocWireless Networks,” Mobile Computing, T. Imielinski and H. Korth, Eds., Kluwer, 1996, pp. 153–81.

[9] J. Broch, D. B. Johnson, and D. A. Maltz, “The Dynamic Source Routing Protocol for Mobile Ad Hoc Networks,” IETF Internet draft, draft-ietfmanet-dsr-01.txt, Dec. 1998 (work in progress).

[10] V. D. Park and M. S. Corson, “A Highly Adaptive Distributed Routing Algorithm for Mobile Wireless Networks,” Proc. INFOCOM ’97, Apr. 1997.

[11] M. S. Corson and A. Ephremides, “A Distributed Routing Algorithm for Mobile Wireless Networks,” ACM/Baltzer Wireless Networks J., vol. 1,no. 1, Feb. 1995, pp. 61–81.

[12] C-K. Toh, “A Novel Distributed Routing Protocol To Support Ad-Hoc Mobile Computing,” Proc. 1996 IEEE 15th Annual Int’l. Phoenix Conf.Comp. and Commun., Mar. 1996, pp. 480–86.

[13] R. Dube et al., “Signal Stability based Adaptive Routing (SSA) for Ad-Hoc Mobile Networks,” IEEE Pers. Commun., Feb. 1997, pp. 36–45.

[14] C-K. Toh, “Associativity-Based Routing for Ad-Hoc Mobile Networks,” Wireless Pers. Commun., vol. 4, no. 2, Mar. 1997, pp. 1–36.

[15] S. Murthy and J. J. Garcia-Luna-Aceves, “Loop-Free Internet Routing Using Hierarchical Routing Trees,” Proc. INFOCOM ’97, Apr. 7–11, 1997.

[16] C. E. Perkins and E. M. Royer, “Ad Hoc On Demand Distance Vector (AODV) Routing,” IETF Internet draft, draft-ietf-manet-aodv-02.txt, Nov.1998 (work in progress). Authors Profile

Vineet Agrawal is having total more than 15 years experience in teacing and industry.Mr. Vineet Agarwal is presently working as a Asst. Director of Rakspal Bahadur College of Engineering & Technology, Bareilly. Author is MCA & M.Tech from Birla Institute of Technology, Mesra Ranchi. Author has worked

in Synthetic & Chemicals Ltd. For four years since 1995 to 1999

as a Sr. Engineer of Computern Application. Mr. Agarwal is the author of number of books. He has written number of books on various topics such as DBMS, Data Structure, Algoritms etc. Mr. Agarwal is also pursuing his Ph.D. in computer scince.He has presented a number of papers in various national conferences.Number of papers have been published in the National & International Journals.Mr Agarwal has also attended various Faculty Development Programe conducted by Infosys and TCS.

Dr. Yahpal Singh is a Reader and HOD (CS) in BIET, Jhansi (U.P.). He obtained Ph.D. degree in Computer Science from Bundelkhand University, Jhansi. He has experience of teaching in various courses at undergraduate and postgraduate level since 1999. His areas of interest are Computer Network, OOPS, DBMS. He has authored many popular books of

Computer Science for graduate and postgraduate level. He has attended many national and international repute seminars and conferences. He has also authored many research papers of international repute.

Manish Varshney received his M.Sc (C.S) degree from Dr. B.R.A. University, Agra, M.Tech. (IT) from Allahabad University and Pursuing PhD in Computer Science. He is working as a HOD (CS/IT) in SRMSWCET Bareilly. He has been teaching various subjects of computer science for more than half a decade. He is known for his skills at bringing advanced computer topics down to

the novice's level. He has experience of industry as well as teaching various courses. He has authored various popular books such as Data Structure, Database Management System, Design and Implementation of Algorithms, Compiler Design books for the technical students of graduation and postgraduation.He has published various research papers in National and International journals. He has also attended one faculty development program organized by Oracle Mumbai on Introduction to Oracle 9i SQL and DBA Fundamental I.

Vidushi Gupta received her B.tech (C.S) degree from Uttar Pradesh Technical University, Lucknow.She is also pursuing M.tech from Karnataka University, She is working as Lecturer ( CS/IT department) in SRMSWCET, Bareilly .She has published a research paper in an International journal. She has also attended one faculty development

program based on the “Research Methodologies”.


78

A Secure Iris Image Encryption Technique Using Bio-Chaotic Algorithm

Abdullah Sharaf Alghamdi1, Hanif Ullah2 Department of Software Engineering,

College of Computer and Information Sciences, King Saud University, Riyadh, Kingdom of Saudi Arabia

[email protected] , [email protected]

Abstract: Due to dramatic enhancement in computers and communications and due to huge use of electronic media, security gains more and more importance especially in those organizations where information is more critical and more important. The older techniques such as conventional cryptography use encryption keys, which are long bit strings and are very hard to memorize such a long random numbers. Also it can be easily attacked by using the brute force attack technique. Instead of using the traditional cryptographic techniques, Biometrics like Iris, fingerprints, voice etc. uniquely identifies a person and a secure method for stream cipher, because Biometric characteristics are ever living and unstable in nature (with respect to recognition). In this paper we used the idea of bio-chaotic stream cipher which encrypts the images over the electronic media and also used to encrypt the images to store it into the databases to make it more secure by using a biometric key and a bio-chaotic function. It enhances the security of the images and it should not be compromised. The idea also gives birth to a new kind of stream cipher named bio-chaotic stream cipher. The paper also describes how to generate an initial key also called initial condition from a biometric string and how to encrypt and decrypt the desired data by using the bio-chaotic function. Keywords: Biometric, stream cipher, bio-chaotic algorithm (BCA), cryptography, key.

1. Introduction

Due to dramatic enhancement in computers and communications and due to huge use of electronic media, security gains more and more importance especially the security of biometric images become a hot issue. Biometric images are mostly used for the authentication system because of there ever living and unstable (with respect to recognition) characteristics. Conventional or traditional symmetric or asymmetric cryptography is limited only to text files but it cannot be used in case of huge files like images and videos. Image encryption techniques are extensively used to overcome the problem of secure transmission for both images and text over the electronic media by using the conventional cryptographic algorithms. But the problem is that it cannot be used in case of huge amount of data and high resolution images [2]. Instead of using the traditional way of cryptography for image encryption we can also use biometric e.g. fingerprint, iris, face, voice etc for the same purpose. The main advantage of a biometric is that it is ever living and unstable characteristics of a human being and it cannot be compromised. However it also suffers from some biometric specific threats and that is the privacy risk in biometric systems. An attacker can interpret a person’s biometric data, which he can use for many illegal operations such is to

masquerade like a particular person or monitor the person’s private data [3]. Similarly some chaos-based cryptosystems are used to solve the privacy and security problems of biometric templates. The secret keys are randomly generated and each session has different secret keys. Thus biometric templates are encrypted by means of chaotic cryptographic scheme which makes them more difficult to decipher under attacks [4]. Moreover some chaotic fingerprint images encryption techniques are also proposed which combines the shuttle operation and nonlinear dynamic chaos system. The proposed image encryption technique provides an efficient and a secure way for fingerprint images encryption and storage [5]. Similarly some new image encryption technique based on hyper-chaos is also proposed, which uses an image total shuffling matrix to shuffle the pixel positions of the plain image and then the states combination of hyper-chaos is used to change the gray values of the shuffled image [6]. In order to improve the security of the images we proposed a better idea which is a new type of algorithm called Bio-Chaotic stream cipher algorithm (BCA) for image encryption which overcomes the problems of some of the algorithms used previously for the same purpose. In this algorithm we used the iris images and extract their features by using the L.Rosa [9] iris feature extraction code. These features are then used to generate the initial condition for the secret key using the Hamming Distance technique, which is then Xored to the iris extracted features to generate another secret key called biometric key. This biometric key is then used in the chaotic function to generate the bio-chaotic stream cipher for further encryption. The rest of the paper is organized such that section 2 consists the related work of the paper. Section 3 will show the basic working and idea of the BCA. Section 4 presents the graphical representation of the key generation process and logistic map for the algorithm. Section 5 shows some mathematical comparisons with other algorithms. Finally section 6 draws a conclusion.

2. Related work

The same work is carried out in our conference paper already published. The same algorithm is used for the encryption of the Iris images. In this paper we elaborate the algorithm with more detail and add some new features to the existing proposed system [19]. The work that we seen relevant to our work is that of Haojiang Gao, Yisheng Zhang, Shuyun Liang and Dequn Li which proposed a new chaotic algorithm for image


79

encryption[2]. In this paper they presented a new nonlinear chaotic algorithm (NCA) which uses power function and tangent function instead of linear function. The experimental results demonstrated in this paper for the image encryption algorithm based on NCA shows advantages of large key space and high-level security, while maintaining acceptable efficiency [2]. Similarly the work done by Song Zhao, Hengjian Li, and Xu Yan for the security and Encryption of fingerprint images is more relevant to our work [5]. In this paper they proposed a novel chaotic fingerprint images encryption scheme combining with shuttle operation and nonlinear dynamic chaos system. The proposed system in this paper shows that the image encryption scheme provides an efficient and secure way for fingerprint images encryption and storage [5]. Also the work done by Muhammad Khurram Khan and Jiashu Zhang for implementing templates security in remote biometric Authentication systems seems relevant to us [4]. In this paper they presented a new chaos-based cryptosystem to solve the privacy and security issues in remote biometric authentication over the network. Experimental results derived in this paper shows that the security, performance and accuracy of the presented system are encouraging for the practical implementation in real environment [4]. Similarly a new image encryption technique was introduced by Tiegang Gao and Zengqiang Chen in their paper based on the image total shuffling matrix to shuffle the position of the image pixels and then uses a hyper chaotic function to complex the relationship between the plain image and the cipher image. The suggested image encryption algorithm has the advantage of large key space and high security [6]. Moreover a coupled nonlinear chaotic map and a novel chaos-based image encryption technique were used to encrypt the color images by Sahar Mazloom and Amir Masud Eftekhari-Moghadam in their paper [10]. They used the chaotic cryptography technique which is basically a symmetric key cryptography with a stream cipher structure. They used the 240 bit long secret key to generate the initial

condition and to increase the security of the proposed system [10].

3. Proposed System Bio-Chaotic Algorithm (BCA)

The basic idea of the algorithm is such that we took an iris image and extract its features by using L.Rosa code [9]. L. Rosa used a code to generate a binary pattern from the given iris image. The binary pattern is further divided into small blocks of binary data to make the process simplified, because it is very difficult to encrypt the binary pattern of hundreds of thousands of bits at once. In our case we made each block of 128 bits to make it simpler and to encrypt each block easily. A random block is then selected to create the initial condition for the secret key. The random selection of the block is preferred because of the attackers, so that no one can easily understand that which block is selected for the initial condition. At the transmission time of the image the bits of this random selected block is encrypted by using Quantum Encryption Technique [8]. Quantum encryption uses light particles, also call photons instead of bits at communication time. A photon can have one of the four orientations or shapes, 450 diagonal, -450 diagonal, horizontal or vertical. Each of these represents a bit, - and / represents a 0, while | and \ represents a 1[8]. Fig 1 presents the block diagram of the proposed bio-chaotic algorithm. The basic steps of the algorithm are as follows. I. Generation of the initial condition from the randomly

selected block taken from the binary pattern of the iris image. The technique used to create the initial condition is that of Hamming Distance i.e.

Where n=1, 2, 3, 4…... Some other techniques can also be used for the same purpose like

Figure 1. Block Diagram of Bio-Chaotic Algorithm


80

II. This initial condition is then converted into secret key by using the LFSR method. An LFSR of length n over a finite field Pq consist of n stages [an-1,an-2,an-3,……..,a0] with ai Є of Pq, and a polynomial

III. The secret key and iris template is then Xored in parallel to generate the biometric key by using the equation,

IV. This biometric key is further Xored with different

blocks of the iris template (divided into blocks of 128 bits/block) which encrypts the image in such a way that no intruder or attacker can easily decrypt the image.

V. To make the algorithm stronger and more secure we add the chaotic function to the biometric key and apply it over the iris image which encrypts it in a more secure way. We use the following logistic equation [4].

Where n=1, 2, 3… is the map iteration index and r is the value taken from the algorithm. On the basis of equation 4 we generate the logistic map for different values of the algorithm the detail of which will be given in the next section.

3.1 Decryption Process

The decryption process of the used image is carried on by the same way using the same key used for the encryption process but in the opposite direction i.e. the ciphered image is Xored with the biometric key to get the image back in its original form. The receiver will first decrypt the randomly selected block by using the same technique used for the encryption process i.e. Quantum Decryption technique [8]. After decrypting the selected block the receiver will generate the initial condition with the same procedure used for the encryption process and will decrypt the image. The equation used for the decryption process is as follows.

It shows the Exclusive OR operation.

4. Experimental Analysis of the Algorithm

In order to evaluate and check the performance of the proposed algorithm i.e. Bio-chaotic algorithm we took iris images from one of the renowned database CASIA (Chinese Academy of sciences and institute of Automation) [11]. The database contains a lot of iris images taken from different people eyes. In our case we use 2 or 3 of the iris images from this database to carry out our experimental process. These images are shown in fig.2. The algorithm is analyzed and tested by using different values for x where x is any real value between 0 and 1. Some of the logistic maps based on the experimental analysis performed over sample and encrypted iris images

are included in this section. The logistic maps are derived on the basis of the following mathematical function.

On the basis of the above equation we generate different logistic maps using different values. Fig.3 and 4 shows the statistical correlation curves of the sequence. By observing the maps carefully it’s clear that even changing in a small part of the value the whole map become different. Fig.5 shows the encrypted images by using different chaotic values. From the figure it is clear that how strong the encryption process is that by changing even a small part of the value the image become more and more invisible. Similarly the decryption process is more sample as like the encryption by just Xoring the Ciphered image with the key and we will get the original image.

Figure 2. iris images used for experiments

Figure 3. Logistic map when value= 0.54000000000001

0 20 40 60 80 100 120 0

0.5

1

0 20 40 60 80 100 120 0

0.5

1

Size of Biometric template

Real value b/w 0 and 1 Logistic Map for Cipher and Decipher date


81

Figure 4. Logistic map when value= 0.58000000001

Figure 5. (a) Encrypted image at value= 0.580000000000

Figure.5. (b) Encrypted image at value= 0.7000000000001

Figure 5. (c) Encrypted image at value= 0.9800000000001

Table 1: Avalanche effect of the BCA

5. Statistical Analysis of Bio-Chaotic Algorithm (BCA)

In this section statistical analysis and mathematical observations like Avalanche effect, confusion and diffusion, and entropy of the proposed algorithm are mentioned.

5.1 Avalanche Effect The Avalanche effect refers to a desirable property of the cryptographic algorithms. The Avalanche effect is evident if, when an input is changed slightly (for example, flipping a single bit) the output changes significantly (e.g., half the output bits flip). In the case of quality block ciphers, such a small change in either key or the plain text should cause a drastic change in the cipher text. In our case the Avalanche effect of the proposed system is determined by using the following mathematical equation [12].

Where percentPC (percent difference between plain image and ciphered image) could be found out by using the equation

Where Acc is basically an Accumulator and it could be find out by

In equation 8 DiffPC means the difference between plain image and ciphered image and it is basically the Xor operation between plain image and cipher image. Similar methodology is used to find out the avalanche effect between plain image and key and ciphered image and key. The results of the above equations are tabulated in table1. The table shows that the Avalanche effect between the plain image and ciphered image, and plain image and key is less than 50 percent that is a more desirable value for any

NO AvalanchePC Effect for BCA

AvalanchePK Effect for BCA

AvalancheCK Effect For BCA

1 48.8758 % 47.9530 % 52.9465 %

0 20 40 60 80 100 120 0

0.5

1

0 20 40 60 80 100 120 0

0.5

1

Size of Biometric

Real value b/w 0 and 1 Logistic


82

algorithm. Similarly the Avalanche effect between ciphered image and key is round about 50 percent which is slightly bigger than the rest of the two, but again it is a desirable value for our proposed algorithm. 5.2 Confusion and Diffusion Confusion and diffusion are the two properties of the operation of a secure cipher. Confusion refers to making the relationship between the key and the cipher text as complex and as involved as possible. Diffusion refers to the property that redundancy in the statistics of the plain text is dissipated in the statistics of the cipher text [12]. Confusion and diffusion are the same properties like Avalanche effect which is elaborated in the previous section. The confusion and diffusion of the proposed algorithm is round about 49%, which shows the strength of the proposed system. 5.3 Entropy

Entropy is a measure of the uncertainty or randomness associated with a random variable. It is basically a measure of the average information content one is missing when one does not know the value of the random variable [12]. Entropy can be found by using the equation

By using the above equation we found the entropy of our proposed system which is round about 127.3. The values show better uncertainty and randomness of bits in the algorithm. The probability of each bit is 0.5. The entropy will be high if there is more randomness in the bits used in the ciphered image. Table 2 shows the entropy of our proposed system.

Table 2: Entropy of Bio-chaotic Algorithm

5.4 Histogram of the Images

Figure 6 shows the histogram for the plain and encrypted or ciphered images used in the bio-chaotic algorithm.

0 50 100 150 200 250

0

1000

2000

3000

4000

5000

6000

7000

8000

Figure 6. (a) histogram of the plain image

Bio-chaotic Algorithm Entropy(H(X))

1 64.67


83

0 50 100 150 200 250

0

5000

10000

15000

Figure 6. (b) histogram of the ciphered image

6. Conclusion This paper presents a new and novel idea for the encryption and decryption of the iris images. The proposed algorithm called the Bio-Chaotic Algorithm (BCA) takes an iris image and with the help of L.Rosa code generates the iris features or the binary bits pattern for the image. This binary bits pattern is then divided into small blocks of bits to simplify the process. Each block is that of 128 bits long. Then a random block is selected from all these blocks to create the initial condition. This initial condition is then passed from the LFSR to generate the secret key. A secret key of 128 bits is generated from the result of the LFSR. This secret key is then used for the encryption of the iris image. A Quantum encryption technique is also used to encrypt the randomly selected block, so that no one can easily attack the block used for the generation of the secret key. The same procedure is then used at the receiver end to decrypt the iris image. Chaotic function is used to make the algorithm more secure and make the process of the encryption and decryption more complex. Experimental and statistical analysis of the algorithm shows that the algorithm is stronger and more secure and can be used for the practical implementation of the iris images encryption. 7. Future Work In the future we would like to use the same technique for the encryption of fingerprint images. Also we would like to use a block size more than 128 bits to make the algorithm stronger and more secure. References [1] Arroyo David, Li Chengqing, Li Shujun, Alvarez Gonzalo,

Halang A. Wolfgang,” Cryptanalysis of an image encryption scheme based on a new total shuffling algorithm”, Elsevier , Science Direct, Volume 41, Issue 5, 15 September 2009, Pages 2613-2616

[2]Haojiang Gao, Yisheng Zhang , Shuyun Liang , Dequn Li, “A new Chaotic Algorithm for image Encryption”, Elsevier , Science Direct , Aug 2005.

[3] Andrew Teoh Beng Jin, David Ngo Chek Ling, Alwyn Goh, “ Biohashing : two factor authentication featuring fingerprint data and tokenized random number “ April 2004,”The Journal Of The Pattern Recognition Society “ , Elsevier , April 2004.

[4] Muhammad Khurram Khan, Jiashu Zhang, “Implementing Templates Security in Remote Biometric Authentication Systems”, IEEE Conf. Proceedings on CIS’06, China, pp. 1396-1400, Vol.2, 2006.

[5] Song Zhao, Xu Yan,”A secure and efficient fingerprint images encryption scheme” Proceedings of the IEEE, 2008, pp- 2803-2808.

[6] Gao Tiegang, Chen Zengqiang,” A new image encryption algorithm based on hyper-chaos” Elsevier, Science Direct, Physics Letters A, Volume 372, Issue 4, p. 394-400, 2007.

[7] Muhammad Khurram Khan, Jiashu Zhang, “Improving the Security of ‘A Flexible Biometrics Remote User Authentication Scheme’”, Computer Standards and Interfaces (CSI), Elsevier Science UK, vol. 29, issue 1, pp. 84-87, 2007.

[8] T Morkel 1, JHP Eloff,” Encryption Techniques: A Timeline Approach”, Information and Computer Security Architecture (ICSA) Research Group Department of Computer Science University of Pretoria, 0002, Pretoria, South Africa

[9] Iris code by Luigi ROSA, L'Aquila ITALY (19600bits)”http://www.advancedsourcecode.com/irisphase.asp

[10] Mazloom Sahar, Eftekhari-Moghadam Masud Amir”, Color image encryption based on Coupled Nonlinear Chaotic Map”, the journal of Chaos, Solitons and Fractals 42 (2009) 1745–1754, ELSEVIER, 2009.

[11] CASIA Iris Database. [Online March, 2006] http://sinobiometrics.com.

[12] Shannon, C. E., “A Mathematical Theory of Communication,” Bell System Technical Journal, July 1948, p.623.

[13] Yao-Jen Chang, Wende Zhang, and Tsuhan Chen, “Biometrics-Based Cryptographic Key Generation” 2004 IEEE, USA.

[14] Ren Honge, Shang Zhenwei, Wang Yuanzhi , Zhang Jian , “A Chaotic Algorithm of Image Encryption Based on


84

Dispersion Sampling” The eight International conference on Electronic Measurement and Instruments” 2007 IEEE.

[15] Shenglin Yang, Ingrid M. Verbauwhede,”Secure Fuzzy Vault Based Fingerprint Verification System”, 2004 IEEE.

[16] The MathWorksTM Accelerating the pace of engineering and science. “www.mathworks.com” Date accessed: 2 Feb 2009

[17] Eli Biham, Louis Granboulan, Phong Q. Nguy “Impossible Fault Analysis of RC4 and Differential Fault Analysis of RC4” Computer Science Department, Technion – Israel Institute of Technology, Haifa

[18] J. Daugman,”High confidence visual recognition of persons by a test of statistical independence “, IEEE Transactions on Pattern Analysis and Machine Intelligence vol.15, 1993, pp.1148-61.

[19] Alghamdi S. Abdullah, Ullah Hanif, Mahmud Maqsood, Khan K. Muhammad., "Bio-chaotic Stream Cipher-Based Iris Image Encryption," cse, vol. 2, pp.739-744, 2009 International Conference on Computational Science and Engineering, Canada.

[20] J.G. Daugman, “Uncertainty Relation for Resolution in Space, Spatial Frequency, and Orientation Optimized by Two-Dimensional Visual Cortical Filters,” J. Optical Soc. Amer., vol. 2,no. 7, pp. 1,160-1,169, 1985.

Authors Profile

Dr. Abdullah Alghamdi is a full time associate professor, SWE Department, College of Computer and Information Sciences, KSU. He holds a Ph.D. in the field of Software Engineering from the department of computer science, Sheffield University, UK, 1997. He obtained his M.Sc. in the field of software development technologies from the UK in 1993. In the academic year 2004/5 he worked as a visiting

professor at School of IT and Engineering, University of Ottawa, Ottawa, Canada, where he conducted intensified research in Web Engineering as part of his Post-Doc program. He recently published a number of papers in the field of Web engineering methodologies and tools. Dr. Abdullah worked as a part-time consultant with a number of governmental and private organizations in the field of IT strategic planning and headed a number of IT committees inside and outside KSU. Currently he is chairing the Software Engineering Department at KSU and part time consultant at Ministry of Defense and Aviation.

Hanif Ullah received the BIT (Hons) and MSc. Degree in Information Technology from Iqra University Karachi in 2004 and Quaid-e-Azam University Islamabad, Pakistan in 2007 respectively. In January 2008, He joined King Saud University, Saudi Arabia as a Research Assistant and start working on Network and Information security related topics. Currently He is working as a Lecturer in the Department of Software Engineering, College of

Computer and Information Sciences, King Saud University, Saudi Arabia.


85

An ASes stable solution in I-Domain G. Mohammed Nazer1 and Dr.A.Arul Lawrence Selvakumar2

1Asst.Pofessor & Head, Dept of MCA, IFET College of Engineering,

Villupuram, India. [email protected]

2Professor & Head, Dept of CSE & IT, Kuppam College of Engineering,

Kuppam, India. [email protected]

Abstract: Routers on the Internet use an interdomain routing protocol called the Border Gateway Protocol (BGP) to share the routing information between the Autonomous Systems (ASes). These ASes defines local BGP policies that lead to various routing anomalies like BGP divergence. In this paper, we close a long-standing open question of Griffin and Wilfong, by showing, for any network structure, if there exists two stable routing outcomes, then there is a possibility of BGP oscillations. Our results provide the first non-trivial necessary condition for BGP safety – uniqueness of the stable routing outcome.

Another question, which is closely related to BGP, is how long it will take to converge to a stable routing outcome. We also address this by analyzing a formal measure of the convergence time of BGP for the policies presented by Gao and Rexford. Even for the restricted class of preferences, we prove that (i) the convergence time is linear in the size of network (ii) BGP’s running time cannot be more than (roughly) twice the length of the longest customer-provider chain in the network.

Keywords: BGP, Border Gateway Protocol, Interdomain routing, network security, routing, networks, routing protocols, BGP safety.

1. Introduction

BGP is the de facto protocol enabling interdomain routing in the Internet. The task of Interdomain routing is to establish routes between the administrative domains which are called as Autonomous Systems (ASes) in the Internet. Global routes are formed from the local decisions that are based on the private routing policies. These routing selections are communicated by the ASes to the neighboring ASes. Persistent routing oscillations are formed due to the lack of global coordination between the local routing policies. BGP safety – Unique stable routing outcome: The main contribution in this paper is showing that BGP safety necessitates the existence of a unique stable solution. This is the result that closes the long-standing open question first posed by Griffin and Wilfong [8]. To be more precise, Two stable solutions in a network implies that the network is unstable that lead to oscillations. To analyze the BGP dynamics in a more simplified form, we use a more convenient structure, called state-transition graph. The state-transition graph, not only a useful conceptual tool for evaluating and designing various network configurations but also assist in detecting the potential routing oscillations and how to debug them.

BGP convergence time analysis: How long it takes BGP to converge to a stable routing outcome? This is another question, which is closely related to BGP. To answer this question, we require a formal definition of measuring the convergence rate, as the Internet is asynchronous.

We analyze the BGP convergence time in particular, Internet-like settings. In this Gao and Rexford settings, every pair of neighboring ASes can have a business relationship or a peering relationship, which causes natural constraints on the ASes’ routing policies.

However, our first result is negative. We show that, even for the restricted class of preferences, there are instances such that the convergence rate of BGP is linear in the size of the network. Specifically we show that in a network with n nodes, it takes n phases to converge. We also prove that the lower bound is tight: BGP is always guaranteed to converge in n time steps. As there are thousands of ASes in today’s Internet, the linear bound does not signify well. However, one would expect BGP to converge at a much quicker rate in practice as ASes’ routing policies are local in the sense that they are not influenced by ASes that are far away. We prove that the number of phases required for convergence is bounded by approximately twice the depth of customer-provider hierarchy.

2. A formal Model

2.1 BGP dynamics

Network model and its policies: In our model, we define a network by an AS graph G = (N, L), where N represents the set of ASes, and L represents number of physical communication links between ASes. N consists of n source-nodes {1,…,n} and a unique destination node d. P i denotes the set of all simple non-cyclic routes from i to d in G. Each source-node i has a ranking function ≤i, that defines a strict order over Pi (that is i has strict preferences over all routes from i to d). We allow ties between two routes in P i only if they share the same first link (i,j). The routing policy of each node i consists of ≤i and of i’s import policy and export policy.

i’s import policy dictates which set of routes Im(i) ⊆ Pi i is willing to send traffic along. We assume that ø ≤i Ri for any route Ri ∈ Im(i) (i prefers any route in Im(i) to not getting a route at all) and that R’

i ≤i ø for any route R’I∉


86

Im(i) (i will not send traffic at all rather than send traffic along a route not it Im(i)).

i’s export policy dictates which set of routes Ex(i,j) ⊆ P i i is willing to announce to each neighbor j.

Update Messages and Activation Sequence: Basically the BGP belongs to a family of routing protocols named path-vector protocols. In this model, there are two kinds of actions that an active node may carry out potentially change the global routing state:

• A node i may select a route or change its selected route from the routes to given destination d that it currently believes to be available.

• A node j may send an update message to a neighboring node i, informing i of the route that j is currently using to destination d. The update message is assumed to be delivered immediately (without propagation delay), and is immediately reflected in updated beliefs that i has about j’s route.

The selection and update actions can occur at arbitrary times. In particular, note that the update messages are not required to be sent at a given time interval or whenever j’s route changes. It is easy to show that a stable solution is always in the form of a tree rooted in d. Further, the import and export policies can be folded into the routing policies, by modifying the preferences so that paths that are filtered out have the lowest possible value.

2.2 The State-Transition Graph

In this subsection, we describe the state transition graph –

a tool that we use to analyze the convergence of BGP on different instances.

The state-transition graph of an instance of BGP is defined as follows: The graph consists of a finite number of states, each state s is represented by an n-dimensional

forwarding vector of routing choices rs = (rs1,…,rs

n), and n ∗

n knowledge matrix Ks = {ksij}i,j. rs

i specifies the identity of the node to which node i’s traffic is being forwarded, and ks

ij specifies the (loop-free) route that node i believes that its neighboring node j is using. We define ks

ij = NULL when j is not a neighbor of i; any knowledge that i has about non-neighboring nodes’ routes is irrelevant to i’s route selection and advertisement decisions. We assume, naturally, that node i knows who it is forwarding traffic to: rs

i must be the first hop in ks

ij. We allow two types of atomic actions that lead to transitions from s to s’: • Route transition – Route selection actions: Informally, a

route transition arises when a node I updates its selected route by picking its favorite route from its current knowledge set of routes used by its neighbors. Formally, there is an i-route transition from state s to state s’ if there is a node i such that: The forwarding vector in s’ is identical to the forwarding vector in s with the possible exception of i.

• Knowledge transition – Informally, a knowledge transition is an update message sent from a specific

node i to a neighboring node j announcing the route that i believes it is sending traffic along. Formally, there is a ji-knowledge transition from state s to state s’ if there is a node i and a neighboring node j such that: The forwarding vectors in the two states are identical. The knowledge matrix in s’ is identical to the knowledge matrix in s with the exception of i’s belief about j, and ks’

ij = ksij. In other words, i learns of the

route that j currently believes it is using. This definition reflects the restricted asynchrony of our

dynamic model. We can phrase this restriction equivalently as: Update messages can be delayed in transit, but when they are delivered, a fresh update message from the same sender is delivered immediately (and thus overrides the delayed update.) Thus, the state description does not have to include messages in transit. Stability and Oscillations in the State-Transition Graph:

A stable state is one in which the nodes forward traffic along a stable solution, and have complete and accurate knowledge about their neighbors’ routes. We want to prove the existence of potential BGP oscillations in the state transition graph. In many cases, oscillations occur only for specific timings of asynchronous events. In particular, starting at any given point of time, every node eventually updates its route selection if its knowledge of routes has changed, and every node eventually receives update messages from each neighbor that has changed a route.

Further, in a given router, there can only be a finite number of other activations taking place between subsequent routing selections or updates. It is for this reason, we look for oscillations that can arise through a fair activation sequence. An infinite activation sequence σ said to be fair if each transition in A appears infinitely often in σ. A fair cycle in the state-transition graph is a finite cyclic path that does not contain a sink, such that every action in A is taken at least once in each traversal of the cycle.

2.3 Implications for the evaluation model of Griffin

We modify the dynamic evaluation model of Griffin in two ways:

• Update messages are not delayed, instead, arrive immediately to the destinations.

• In BGP execution, it is not necessary that a node inform a neighboring node of every new route it changes, rather it is enough if it announces once in a while.

3. Two stable solutions leads to BGP oscillation

In this section we prove our main result, that if there are two stable solutions then the network is unstable in the sense that persistent route oscillations are possible.

Theorem: If the AS graph G contains two stable solutions, then there is a fair activation sequence under which BGP will oscillate on G. That is, two stable solutions imply that the network is unstable, in the sense that it could plausibly lead to persistent route oscillations. Therefore, to achieve


87

BGP stability, the network must have a unique stable solution.

The intuition behind our proof is as follows. In the state-transition graph, each stable state will have a corresponding “attractor region”: a subset of states (possibly just the stable state itself, or much larger) that, once reached, we can be certain that the system will ultimately converge to the stable state. We can visualize the state-transition graph as a map, with each attractor region a different color – red, blue, etc. However, there will also be some states that do not lie in any one attractor region, because different evolutions from that state could lead to different stable states. We label these states with a distinct color – purple, say – and show that the Zero state must belong in this subset.

The key to the proof is showing that, starting from any purple state, we can find a fair activation sequence that ends at another purple state. We use the properties of route selection and update actions to show that we can swap the order of any two consecutive activations, perhaps repeating one of them, and achieve the same result as the original order. Thus, it is not possible that any given activation a leads to a red state in the original order, but leads to a blue state in the perturbed order. Using this, we show that we can add each activation while staying within the purple region. As the graph is finite, this implies the existence of a fair cycle. If an instance of BGP results in a state- transition graph (for a given destination) that has a fair cycle, we will infer that there is a plausible sequence of route selections and updates that will cause BGP to oscillate.

4. BGP’s convergence Rate

In this section, we handle the question of how long BGP takes to converge to the unique stable solution. BGP is an asynchronous protocol, and individual messages may be lost or delayed arbitrarily. As we cannot assume a bound on the actual elapsed time of a single message, any model of convergence “time” needs to define a unit of time measurement that remains meaningful in this asynchronous setting. Let us consider the following definition:

Definition: A BGP phase is a period of time in which all nodes get at least one update message from each neighboring node, and all nodes are activated at least once after receiving updates from their neighbors.

We analyze the number of BGP phases it requires for the network to converge. The underlying principle in this definition is that, although it is difficult for the analyst to assert numerical bounds on the update frequencies at different nodes, it is reasonable to expect that all nodes are updating at similar timescales. The definition of phases admits asynchrony, thus capturing the realistic possibility that different sequences of update activations can lead to different transient behavior. At the same time, by tying the unit of measurement to the slowest node’s update instead of a fixed time unit (or the fastest update), we avoid pathological worst-case time bounds that are only attained if, for example, one node’s update cycle is measured in years instead of seconds or minutes.

How many consecutive phases does it take BGP to converge to a stable solution in the worst case? Routes are propagated through the network one hop at a time, so the best we can hope for is a time proportional to the length of

the longest route in the stable solution. It is easy to construct instances with n nodes in which there are routes of length Ω(n). However, these instances are unnatural; currently, Internet routes tend to be much shorter than this. For this reason, we focus on bounding the BGP convergence time on Internet-like graphs.

Example: The graph in Figure 1 depicts a network with n nodes, and a destination node d. Node 1 prefers to go directly to d. Any other node i prefers the route i → i−1 → d over the direct route i → d. All routes of length greater than 2 are less desirable to any node. This set of path preferences is compatible with the Gao-Rexford constraints for the following set of customer-provider relationships: 1 is a customer of 2, 2 is a customer of 3, etc.; and, additionally, d is a customer of every other node.

In each phase, initially all update messages go through, and then all nodes are activated. In the first phase, only node 1 will change its routing choice and will route to d. In the next phase, only node 2 will change its routing choice and will route through 1. Then node 3 will change to route through d and so on. The network will eventually converge to the routing outcome in which all odd nodes route directly to d and all even nodes route the rough their counter-clockwise neighbor.

We prove that this bound is tight for the class of instances that satisfy the Gao-Rexford conditions. In fact, we prove a slightly stronger result: The following proposal shows that our bound on BGP’s convergence rate is tight on the larger class consisting of all instances in which the “No Dispute Wheel” condition of [3], [5] holds.

Proposal: If “No Dispute Wheel” holds then BGP’s convergence rate is at most n phases.

Proof: Let us assume that indeed the “No Dispute Wheel” condition holds in a network graph G with a destination node d. At every phase, one of the nodes of the graph converges to a route that will not change from that point on. The first node that converges in the first phase is the destination node d, that has the empty path, and announces that path to its neighbors. We now show that there must exist a node in the network that is a direct neighbor of the destination d and that its most preferred path is going directly to d.

To see that this is indeed the case, pick an arbitrary node v, look at its most preferred path to the destination. This path goes through a neighbor of d right before it reaches d. We shall denote this neighbor by v1. Now, consider the most preferred path of node v1, and the closest node to d on that path that we shall denote by v2. In this manner we define the nodes vi for i = 1, 2, 3, ... At some point, nodes in the sequence we defined must start repeating. If only one node


88

repeats infinitely then this node must have a direct route as its most preferred path, and we are done. Otherwise, the sequence of repeating nodes vk , vk+1 , . . . , vk+l (for some k, l) forms a dispute wheel: each node prefers to go through the next one in the sequence rather than directly to d. This contradicts our assumption. Therefore, there exists a node that prefers to go directly to d over any other path. It will choose this path to d on the second phase, send update messages to its neighbors, and will never again change its path (since no path will be better).

We now continue to follow the convergence process, and observe that at any phase, there must exist a node v that converges to its most preferred route given the route of the nodes in the system that have already permanently converged. This node will never again change its path (because unless previous nodes change, it will have no better path, and these previous nodes have also converged). To prove that such a node must exist, we fix the routes of all permanently converged nodes, and pick an arbitrary node v1 that did not converge. We once again define the sequence of nodes v1, v2 , v3 . . . by defining the node vi+1 as the node that is closest to d on the most preferred path of node vi that did not permanently converge. The set of paths from which we select this most preferred path, is the set of paths that are consistent with the nodes that have already permanently converged. Once again, this sequence of nodes must repeat, and since it cannot contain a dispute wheel, it must have only a single repeating node that is the closest node that did not converge on its own most preferred path. In the next phase, this node’s path converges. We have thus shown that if the AS graph contains no dispute wheels, the convergence time of BGP is bounded by the number of nodes in the entire network graph.

5. Conclusion

We studied fundamental questions related to BGP

whether it will converge to a unique stable solution and how long it will take to converge. We proved that, for any network, if there exists two stable routing outcome, then persistent BGP oscillations are possible. So the existence of unique stable routing outcome is a necessary condition for the BGP safe convergence. We also analyzed the worst-case convergence time of BGP on instances that satisfy the conditions mentioned by Gao-Rexford. We proved that the convergence time on a graph with n nodes is Θ(n) in the worst case, but is much smaller in networks with shallow customer-provider hierarchies.

An interesting direction for future research is proposing formal models for addressing these issues and assessing their impact on our necessary condition for BGP safety. First, can we close the gap between our necessary condition and known sufficient conditions for safe convergence? Second, can we develop a compositional theory for safe policies? If we put together two sub networks with unique stable solutions, when does the combination also have a unique stable solution? It would also be valuable to extend the convergence-time analysis to broader classes of preferences, and to characterize the average-case (instead of worst-case) convergence time following a network change.

Answers to these questions could provide network operators with new principles to tradeoff the desire for flexible autonomous policies with the need for global routing efficiency. Finally, there are practical aspects of BGP operations not considered in this paper such as MRAI (Minimum Route Advertisement Interval) and RFD (Route Flap Damping [19]), which play a significant role in BGP convergence [20], [21]. References

[1] K. Varadhan, R. Govindan, and D. Estrin,

“Persistent route oscillations in inter-domain routing,” Computer Networks, vol. 32, no. 1, pp. 1–16, March 2000.

[2] T. G. Griffin and G. Wilfong, “An analysis of BGP convergence properties,” in Proceedings of SIGCOMM 1999.

[3] T. G. Griffin, F. B. Shepherd, and G. Wilfong, “The stable paths problem and interdomain routing,” IEEE/ACM Transactions on Networking, vol. 10, no. 2, pp. 232–243, April 2002.

[4] L. Gao and J. Rexford, “Stable Internet routing without global coordination,” IEEE/ACM Transactions on Networking, vol. 9, no. 6, pp. 681–692, 2001.

[5] L. Gao, T. G. Griffin, and J. Rexford, “Inherently safe backup routing

with BGP,” in 20th INFOCOM. Pistacaway: IEEE, 2001, pp. 547–556.

[6] T. G. Griffin, A. D. Jaggard, and V. Ramachandran, “Design principles of policy languages for path vector protocols,” in SIGCOMM ’03: Proceedings of the 2003 conference on Applications, technologies, architectures, and protocols for computer communications. New York: ACM, 2003, pp. 61–72.

[7] A. D. Jaggard and V. Ramachandran, “Robustness of class-based path- vector systems,” in Proceedings of ICNP’04, IEEE Computer Society. IEEE Press, October 2004, pp. 84–93.

[8] N. Feamster, R. Johari, and H. Balakrishnan, “Implications of autonomy for the expressiveness of policy routing,” in SIGCOMM ’05: Proceedings of the 2005 conference on Applications, technologies, architectures, and protocols for computer comm. New York, NY, USA: ACM Press, 2005.

[9] Sobrinho, “An algebraic theory of dynamic network routing,” IEEE/ACM Transactions on Networking, vol. 13, no. 5, pp. 1160–1173, 2005.

[10]T. G. Griffin and G. Huston, “TRFC 4264: BGP wedgies,” 2005.

[11]L. Subramanian, S. Agarwal, J. Rexford, and R. Katz, “Characterizing the internet hierarchy from multiple vantage points,” INFOCOM 2002. Twenty-First Annual Joint Conference of the IEEE Computer and Comm.Societies. Proceedings. IEEE, vol. 2, pp. 618–627, 2002.

[12] C. Labovitz, A. Ahuja, A. Bose, and F. Jahanian, “Delayed internet routing convergence,” SIGCOMM


89

Comput. Commun. Rev., vol. 30, no. 4, pp. 175–187, 2000.

[13] J. Feigenbaum, R. Sami, and S. Shenker, “Mechanism design for policy routing.” Distributed Computing, vol. 18, no. 4, pp. 293–305, 2006.

[14] H. Karloff, “On the convergence time of a path-vector protocol,” in SODA ’04: Proceedings of the fifteenth annual ACM-SIAM symposium on Discrete algorithms. Philadelphia, PA, USA: Society for Industrial and Applied Mathematics, 2004, pp. 605–614.

[15] T. G. Griffin and G. Wilfong, “A safe path vector protocol,” in Proceedings of IEEE INFOCOM 2000, IEEE Communications Society. IEEE Press, March 2000.

[16] H. Levin, M. Schapira, and A. Zohar, “Interdomain routing and games,” in Proceedings of the 40th ACM Symposium on Theory of Computing (STOC), May 2008.

[17] A. Fabrikant and C. Papadimitriou, “The complexity of game dynamics: BGP oscillations, sink equlibria, and beyond,” in Proceedings of SODA 2008.

[18] G. Huston, “Interconnection, peering, and settlements,” in Internet Global Summit (INET). The Internet Society, 1999.

[19]Z. M. Mao, R. Govindan, G. Varghese, and R. H. Katz, “Route flap damping exacerbates internet routing convergence,” in SIGCOMM ’02: Proceedings of the 2002 conference on Applications, technologies, architectures, and protocols for computer communications. New York, NY, USA: ACM, 2002, pp. 221–233.

[20]E. C. Jr., Z. Ge, V. Misra, and D. Towsley, “Network resilience: Exploring cascading failures within bgp,” in Allerton Conference on Communication, Control and Computing, October 2002.

[21]K. Sriram, D. Montgomery, O. Borchert, O. Kim, and D. R. Kuhn, “Study of bgp peering session attacks and their impacts on routing performance,” IEEE Journal on Selected Areas in Communications, vol. 24, no. 10, pp. 1901–1915, 2006.


90

Biometrics Based File Transmission Using RSA Cryptosystem

Mr.P.Balakumar1, Dr.R.Venkatesan2

1Assistant Professor, Department of Computer Science and Engineering,

Selvam College of Technology, Namakkal, Tamilnadu, India. [email protected]

2Professor & Head, Department of Information and Technology,

PSG College of Technology, Coimbatore, Tamilnadu, India. [email protected]

Abstract: Biometrics gives a lot of methods in high-secure applications while using natural, user-friendly and fast authentication. Most of the implementations of Public key Cryptosystems widely use the RSA algorithm. The RSA algorithm is one of the asymmetric algorithms in which we use two keys private and public. The efforts in this paper merge the biometric concept with the asymmetric cryptography to offer the security for the document sending process in the distributed network. For document sending the sender encrypts the message using the receiver’s public key and for decryption the receiver uses his private key. This system uses the fingerprints as the security-providing medium. This system is developed under Graphical User Interface environment which is very easy to operate by the users. This system is developed using the Java language so that it can be executed on any platform. The design of this system supports both the Internet and Intranet environments. Dynamic key generation process is the main contribution of this work.

Keywords: Cryptography, Biometrics, RSA, DSS, KDC 1. Introduction A biometrics system is a standard method for the identity verification of a human being based on the personal or physical identification of characteristics. The functions of biometric systems are determining, measuring and codification of the unique characteristics of individual persons with one already recorded. In recent years there has been rapid growth in the use of biometrics for user authentication applications because biometric based authentication provides several benefits over knowledge and possession-based methods. General biometric systems consist of the four phases. They are, data collection which includes sensing and pre-processing, signal analysing which includes feature extraction and template generation, storage, and decision making with a matcher as shown in Fig. 1. A secured biometrics system does not change widely over a long time, but a less secure biometrics system is likely to

change with time. For example, the iris-recognition does not change over a human’s lifetime and it is more secure than voice-identification. Uniqueness in human’s biometric is a scale of the variations or differences in the biometric model among the worldwide population. The high-level degree of uniqueness produces more unique identifier. A low-level degree of uniqueness indicates a biometric pattern that is found commonly in the general population. The iris and retina have higher levels of uniqueness than hand, voice and finger printing. The nature of an application helps in determining the degree of strength and uniqueness needed. Living persons distinguish the biometric verification from forensics, which does not involve real-time recognition of a living human being.

Figure 1 General Biometric System. Information sharing is a necessary part of our life. Hence, security of information from mishandling is need. A cryptography mechanism provides a set of data transformations called encryption and decryption to send the data in a secured manner. Encryption is applied to the normal message i.e. the data to be translated is used to


91

produce the code message (encrypted data) which is apart from original data using encryption key. Decryption uses the decryption key to convert code message to original message (the original data). Now, if the Encryption key and the decryption key are same or one can be copied from the other then it is said to be symmetric cryptography.

There is a drawback in symmetric cryptography. That is the sender must send the same key to the receiver through another secured channel. The attacker can capture it and he could find the original secret key. This type of cryptography system can be easily broken if the key used to encryption or decryption is known.

To overcome the drawback present in the symmetric cryptography we moved towards Public Key Cryptography system that was found in 1976 by Whitfield Diffie and Martin Hellman of Stanford University [22]. It uses a set of associated keys one for encryption and another one for decryption. One key, which is known as the private key, is kept top secret by the user and another one key is public key that is distributed to all other users. 2. Security of Biometrics Regular biometrics can help to reduce the problems related with the existing methods of user verification. The hackers will find the weak points in the existing system and attack the existing system accordingly. Unlike key systems, which are able to find the message using brute-force attack, biometric based systems are difficult to crack. The biometric systems need considerably more attempts to breakthrough. Although standard encryption techniques are helpful in many ways to avoid breach of security, there are some new types of attacks are possible. If biometric system is used as a supervised verification tool, there may not be problems, but in a distant unattended application, such as web oriented, e-commerce applications, hackers may have sufficient time to make frequent attempts before being noticed or may even be able to actually break the remote client [8]. 2.1 Comparison to Password Real benefits of biometric systems are that they are much longer in size than a password or phrase key. They vary from hundred bytes to over a megabyte. Usually the message content of such signals is relatively high. It is almost not possible to keep in mind a 2K password and it would take an tediously long time to type in such a password anyhow (particularly with no errors). Fortunately, automated biometrics can offer the security advantages of long passwords while still retaining the speed and simplicity of short passwords. Still, in general smaller amount of them are typically covered, such as dissimilarity is that there is no “fake password” input detector equivalent to the fake biometric.(although perhaps if the password was in some standard dictionary it could be deemed “fake”). Additionally, in a password or token based verification system no effort is made to prevent replay attacks (since there is no difference of the “signal” from one presentation to another). However, in an automated biometric-based verification system, one can go to the extent of checking liveliness of the input signal.

Another significant difference concerns the matching subsystems. A password based method always provides a crispy result. If the password matches, it grants access and otherwise refuses access. However the performance of a pattern detection system in general is dependent relative on several factors such as the quality of input and enrols data along with the basic characteristics of the underlying algorithm. This is typically reflected in a graded overall match “score” between the submitted biometric and a stored reference. In a biometrics-based system, they can purposely set a threshold on the score to directly control the false acceptance and false rejection rates. Inverting this, given a good matching score the system can guarantee that the probability of signals coming from a genuine person is significantly high. Such a calibrated confidence measure can be used to tackle non-repudiation support – something that passwords cannot provide [8]. 3. RSA Algorithm The Rivest, Shamir, Adelman (RSA) scheme is a block cipher asymmetric cryptosystem, in which the Plaintext and ciphertext are integers between 0 and n-1 for some n. A typical size for n is 1024 bits or 309 decimal digits. In RSA system all the users must generate their private key KR={d,n} and kept it in secret and store their public key KU={e,n} in Key Distribution Centre(KDC). The sender receives the receiver’s public key from the KDC and encrypts the message using the receiver’s public key. The receiver uses his private key to decrypt the coded message. The private key is known only to the receiver himself.

3.1 Finger Prints The finger printing biometrics is an automatic digital version of the old ink-and-paper method used for more than a century for recognition, mainly by law enforcement agencies. Some samples of fingerprint images are shown in the Figure.2. The Biometric device involves users placing their finger on a platen for the print to be read. The minutiae are then extracted by the vendor’s algorithm, which also makes a fingerprint pattern analysis. Fingerprint template sizes are typically 50 to 1,000 bytes. The Fingerprint biometrics currently has three main application areas: Large-scale Automated Finger Imaging Systems (AFIS) generally used for law enforcement purposes, fraud prevention in entitlement programs, and physical and computer access.

Figure 2. Sample Fingerprints


92

3.2 Characteristics of Biometrics

Table.1 compares the seven mainstream biometrics in terms of a lot of properties, ranging from how robust and distinct [10] they are to what they can be used for (i.e., identification or verification or verification alone). This table is an effort to lend a hand to reader in categorizing biometrics along important dimensions. Because this industry is still functioning to establish comprehensive standards and the technology is varying rapidly, however, it is difficult to make assessments with which everyone would agree. The table shows an assessment based on consideration with technologists, vendors, and program managers. The table is not proposed to be an aid to those in the market for biometrics; rather it is a guide for the unskilled.

Table.1 Comparison of Mainstream Biometrics

Biometric Identify versus Verify

Robust

Distinctive

Intrusive

Fingerprint Either Medium High Touching

Hand Verify Medium Low Touching

Facial Either Medium Medium 12+ inches

Voice Verify Medium Low Remote

Iris Scan Either High High 12+ inches

Retinal Either High High 1–2 inches

Keystroke Dynamics

Verify Low Low Touching

When comparing ways of using biometrics, half can be used for both the identification and verification, and the remaining can only be used for authentication. In specific, hand geometry has only been used for confirmation applications, such as physical access control and time and attendance verification. Adding to this, voice detection because of the need for staffing and matching using a pass-phrase, is used for verification only. There is considerable changeability in terms of robustness and individuality. Fingerprinting is sort of robust, and, even though it is distinctive, a small proportion of the population has unusable prints, always because of age, genetics, injury, career, spotlight to chemicals, or other occupational hazards. Hand/finger geometry is moderate on the distinctiveness scale, but it is not very robust, while facial recognition is either highly robust or distinctive.

In voice recognition, assuming the voice and not the pronunciation is being measured; this biometric is moderately robust and distinctive. Iris scans are both highly robust because they are not highly vulnerable to routine changes or damages and distinctive because they are randomly formed. At last dynamic signature verification and keystroke dynamics are not robust or distinctive. 4. Problem Statement Even though the RSA algorithm uses the finger printing biometrics system to generate the public key and private key generation there are some problems in that approach. They are:

1. Brute-force attack: The maximum size of the public key and private key obtained by RSA algorithm is 155 digits. It can be captured by a brute force attacker using thousands of machines and it requires three month of computation. {Ref: Journal of Telecommunications and Information Technology. Volume 4/2002. Pages 08-09}.

2. Increased key storage requirement: RSA key storage (private keys and public key) requires significant amounts of memory. So, we have to store the public key and private key in any equipment or in brain.{Ref: Journal of Telecommunications and Information Technology. Volume 4/2002. Pages 41-56}.

3. No Dynamic key generation: There is no dynamic key generation in RSA algorithm. Therefore the user must keep secretly his private key. There is a chance to lose or stolen, forgotten the private key of the RSA algorithm, hence he may lose the data.

5. Proposed Scheme The architecture of the proposed scheme is shown in Figure.3. The client generates the public key and sends to the KDC. On document send process it retrieve the receiver’s public key from KDC and encode the data with aid of generated public key. Then it sends the encoded data to the receiver. While viewing the document it dynamically generates the private key which is used to decode the encoded data. The proposed digital signature algorithm is a version of the RSA algorithm that overcomes the problems in the RSA system. A brute force attacker can able to hack the private key by using every possible combination of the key (i.e. Numeric key). In our system, we use alpha numeric (combination of alphabets and numeric) keys, hence the attacker can not able to obtained the key values easily. The second problem in the existing RSA algorithm is key storage requirement. In our proposed system we generate the private key dynamically. Hence there is no need for key storage requirement. The third problem in the existing system is no dynamic key generation. Normally, by using RSA algorithm they have to generate their public key and private key. Then they have to send the public key to the key


93

distribution centre and keep their private keys secretly with themselves. In our proposed algorithm we generate the public key using the finger print and send that public key to the key distribution centre. While encrypting the data the sender get the public key of the receiver from the key distribution centre and encrypt the data with that public key. To decrypt the ciphertext the receiver requires his private key. At that time of decryption only the receiver will be able to know his private key. This process is called ‘Dynamic private key generation’.

Figure 3. Architecture of Proposed Scheme

6. Key Distribution Centre KDC has a very significant role in the asymmetric key cryptosystem. It receives public key values from the clients and stores in its locale. It is the only authoritative system to distribute the public key values to the requesting users. The KDC application is a server application. The KDC application has two modules. One is key management module and another one is key distribution module. The key management module is mainly for receiving and maintaining the key values. On other hand the key distribution module distribute the public key value based on the client requests.

6.1 Key Management Process The key management module is created to perform the key maintenance process. It has two main tasks. They are the key receive process and key expiry management process. The key receive process is run as a separate thread. The KDC listen all the key value and send a key for the client request. For the receiving process, it needs UDP socket. It does not make any connection with the client application. This module maintains the entire received public key values. The key expiry management module keeps the validity of the key values. KDC automatically removes the key values from the key list if the client application process is terminated. The clients can change their key value and update. So the

existing key value is replaced from the list and the new key value is added into the list. 6.2 Client The client application is designed to hold the document transfer process and the key generation process. The client application is divided into four modules. They are the Key generation module, the sender module, the receiver module and the document view module. The key generation module generates the key from the finger print data. The sender module is used to encode and send the document. The receiver module receives the decoded documents that are sent by the other clients. The document view module maintains received documents after the decoding process the user can view the document.

6.3 Key Generation Module The key generation process is shown in Figure.4. This module is to generate the public key by using the finger print data. The input data is given as an image to the system. This value is to create the key base value that is used to generate the public key value.

Figure 4. Key Generation

The public key value in KDC is stored with its client details. The system supports the JPEG and the GIF image formats. The pixel matrix is constructed using image data. The key base is generated by using the image data matrix values. The system has been implemented as a GUI based application developed in Java. The main menu has three options. They are the key preparation, document list and send process. The key generation windows receive the input for the finger print image file. The generate button is used to initiate the key generation process. The send button is used to start the key transfer process. The key distribution centre is designed to receive and maintain all the public key values. The message sending process is used to transfer a file from one client to other. The message file is encoded before the sending process. All the received messages are listed in the inbox. The user can select the file and perform the view process. The documents are decoded before the view process. The private key value is generated at the time of the decoding

Capture the finger print data

Dot matrix Conversion

Generate key Base

Generate public key

Send the key to KDC

Key Update Client

KDC

Pub Key Generation

Document Send Key

Retrieval

Encode Send

Document View Private Key

Generation

Decode

Display


94

process. The decoded documents are stored in the specified folder 7. Testing and Implementation Testing is the important phase in the system development process. The system is developed as a GUI based application. The system is tested before the implementation process. The system is tested with different testing methods. They are unit test, integration test, system test, validation test and stress test. The system is tested with different network and platform environments. The system uses the image scanner to capture finger print image data. The system is tested in the Intranet environment. In this system each and every modules is tested separately for the unit test. For example the RSA algorithms processes key generation, encode and decode operations are tested with the corresponding modules. The Client application and the Key distribution applications are tested separately. The integration test is performed after all the modules are connected with the main menu. The entire system is tested with all the operations by using a set of finger print values. The stress test is conducted to test the load management strength of the client application and the key distribution centre application. Connecting multiple clients with the KDC tests the key distribution centre stress. In the client application sending a large file to the other client tests the strength. The validation test is performed for all input values. The finger print image availability is checked before the key base generation process. The system is developed to distribute document with security using the biometrics. The system is tested and the results are very good. The implementation of the system is conducted as direct change over mechanism. The new system is directly installed and activated into the action for usage. The system can be implemented in any network environment. The system supports all type of file transfer operations. The system has developed as two applications the key distribution centre and the client application. The key distribution centre application is loaded into a separate machine. The client application can be loaded into all other client machines. All the client application should be configured for the key distribution centre IP address for the key updating and request process. The system now designed to get the fingerprints images from the image file. So the system must be connected with an image scanner. The system can also be connected with the fingerprint image scanner devices. The client application and the key distribution applications can be continuously executed to maintain the connection and message receive process. All the messages are directly received by the client applications. The system requires a lesser amount of hard disk space to store the received and decoded documents. The key distribution centre should be connected with all the client applications. The system can be run with one or more network environments.

7.1 Software Selection

Using Java language under windows platform develops the simulation tool. Java supports multiple platforms, GUI design and network operations. Using the Java language develops the system. Image processing, cryptographic operations, network transmissions and file processing are the major are in the system. Java provides a variety of packages and classes to support all these tasks. The user interface is designed with GUI support. The application is designed to run under any platform. The finger print values are retrieved from image files. The image file data are extracted and converted into pixel matrix. Using the classes such as Image, MediaTracker and PixelGrabber in Java the system does these processing. The Image class is used to convert an image into an object. The Media Tracker and the Pixel Grabber classes are used to support the data extraction and pixel conversion process. These classes are available in the java.awt package. Java provides a separate package JCE for the cryptography process. But the JCE requires the Service Providers for the implementation. In Java cryptography can be implemented in two ways. They are using the JCE with service providers and the other one write the code for the cryptographic algorithms. In this work the second method is applied. The RSA algorithm is implemented by using the java.math package support. RSA requires high bit length data type support. Java provides a class Big Integer to process values in 128 bits. All RSA key generation, encoding and decoding operations are done by using the Big Integer class. The file process and the data transmission process are implemented with the support of java.io and java.net package. All the files are processed using the byte stream classes. The data transmission tasks are done using the TCP/IP support classes in the java.net package. The key distribution centre application is designed using the UDP. The client application transfers the files using the Server Socket and Socket classes. Data gram Socket and Data gram Packet classes are used in KDC process. 8. Conclusion The System is developed to provide security for the file transfer process in distributed environment. Document transmission between the systems that are in the distributed environment is a usual task. The same environment is also shared by a lot of members. So the system should ensure the security of the documents that are transferred. Different cryptographic techniques are used to secure the data. In the recent days, biometrics is used to recognize the users. This work combines the biometrics and cryptography to provide the security for the document transmission process in the distributed environment. Generally passwords and smart cards are used for the security systems. The system uses the biometrics technology as the security-providing medium. This system uses the fingerprints for the security system. Password can be hacked by trial and error basis. But it is not possible to break the biometrics based security system. The system is developed as two applications. They are the key distribution centre application and the


95

client application. The KDC supplies the public key values to the required clients. The client application is designed to handle all the data transfer and security operations. The system uses a designed key base generation algorithm and RSA algorithm. The system is tested with various samples and clients. The performance of the system is very good. The system is tested with different type of file formats. The result shows that the system supports all types of file format. The system stores and distributes the public key values for all clients in the key distribution centre. The system does not require any key storage process for the secret key. In this work using the fingerprint values, the system can generate both the public key and private key. Damages that occurred in the finger print may impact the recovery of the documents. In future the system can be implemented for all type of authentication process such as capillary patterns in the retina, hand geometry, facial characteristics, signature dynamics, voice pattern, and timing of keystrokes. Data compression technique can be used to reduce the content size, process time and transmission time. In future the system will include noise detection and filtering facility for the input process.

References [1] Bruce Schneier, “Applied Cryptography Protocols,

Algorithms” 2nd Edition, Wiley publication. [2] Naughton.P and H.Schildt, “Java 2: The Complete

Reference” , McGraw-Hill,1999 [3] William Stallings, “Cryptography and Network

Security Principles and practice”, 2nd Edition, Prentice Hall, Upper Saddle River.

[4] Anil Jain, Lin Hong, Sharath Pankanti, and Ruud Bolle, “An Identity Authentication System Using Fingerprints” Department of Computer Science, Michigan State University East Lansing.

[5] James L. Wayman, “Biometrics Identification” , Communications of the ACM, February 2000.

[6] Katrin Franke, Javier Ruiz-del-Solar, Mario, “Soft-Biometrics: Soft-Computing for Biometric-Applications” Dept of Pattern Recognition, Fraunhofer IPK, Berlin, Germany.

[7] Nalini K. Ratha, Jonathan H. Connell, and Ruud M. Bolle J, “An Analysis of Minutiae Matching Strength” Watson Research Center.

[8] Rowley. T, “Silicon Fingerprint Readers: A solid state approach to biometrics” , Proc. of the Card Tech / Secure Tech, Orlando, Florida, May 97.

[9] Schneier.B, “The uses and abuses of biometrics” . Communications of the ACM, August 1999.

[10] Schneir.B, “Security pitfalls in cryptography” , Proc. of Card Tech /Secure Tech, Washington D.C., April 98.

[11] Wong C K and Lam S S, “Digital Signatures for flows and multicasts”, IEEE/ACM Transaction and Networking”, August 1999.

[12] www.rand.org/publications/MR/MR1237/MR1237.appa.pdf

[13] www.mit.bme.hu/events/minisy2003/papers/orvos.pdf

[14] http://rpmfreelancer.no ip.com:8080/duncan21/biometrics/finger.html

[15] www.cost275.gts.tsc.uvigo.es/presentations/COST275_Jain.pdf

[16] www.research.ibm.com/ecvg/pubs/sharat-proc.pdfM. Wegmuller, J. P. von der Weid, P. Oberson, and N. Gisin, “High resolution fiber distributed measurements with coherent OFDR,” in Proc. ECOC’00, 2000, paper 11.3.4, p. 109.

[17] R. E. Sorace, V. S. Reinhardt, and S. A. Vaughn, “High-speed digital-to-RF converter,” U.S. Patent 5 668 842, Sept. 16, 1997.

[18] The IEEE website. (2002) [Online]. Available: http://www.ieee.org/

[19] M. Shell. (2002) IEEE Transaction homepage on CTAN. [Online]. Available: http://www.ctan.org/tex-archive/macros/latex/contrib/supported/IEEEtran/

[20] W.Diffie and M.Hellman.” New Directions in Cryptography”. IEEE Transaction on Information Theory.IT-22(1978).472-492.

Authors Profile

Mr.P.Balakumar received the B.E. and M.E. degrees in Computer Science and Engineering from PSG College of Technology, Coimbatore, in 1997 and Anna University, Chennai in 2004 respectively. During 1999-2001, he worked as Lecturer in PSG College of Technology in

Coimbatore. Later during 2003-2008, he worked as Lecturer & Assistant Professor in AMS Engineering College, Namakkal. He now with Selvam College of Technology, Namakkal, Tamilnadu, India as Assistant Professor in Department of Computer Science and Engineering.

Dr.R.Venkatesan was born in Tamilnadu, India, in 1958. He received his B.E (Hons) degree from Madras University in 1980. He completed his Masters degree in Industrial Engineering from Madras University in 1982. He obtained his second Masters degree MS in Computer and Information Science from University of Michigan, USA

in 1999. He was awarded with PhD from Anna University, Chennai in 2007. He is currently Professor and Head in the Department of Information Technology PSG College of Technology, Coimbatore, India. His research interests are in Simulation and Modeling, Software Engineering, Algorithm Design, Software Process Management.


96

Artifact Extraction and Removal from EEG Data Using ICA

Ashish Sasankar1, Dr. N.G Bawane2 ,Sonali Bodkhe3 ,M.N.Bawane4

1 G.H.Raisoni College of Engineering

CRPF Gate 3 , Digdoh hills Hingna road Nagpur (India) [email protected]





4 Govt Polytechnic,Nagpur(India) [email protected]

Abstract: The Independent Component Analysis (ICA) has emerged as a novel and promising new tool for performing artifact corrections on EEG data. In this paper, ICA is used to perform artifact correction on three types of artifacts namely, frontal (eye), Occipital (rear-head), and muscle. EEG is composed of electrical potentials arising from several sources. Each source (including separate neural clusters, blink artifact or pulse artifact) forms a unique topography onto the scalp – ‘scalp map‘. Scalp map may be 2-D or 3-D.These maps are mixed according to the principle of linear superposition. Independent component analysis (ICA) attempts to reverse the superposition by separating the EEG into mutually independent scalp maps, or components.

Keywords: EEG, Independent Component Analysis(ICA), BCI

1. Introduction EEG stands for Electroencephalogram. It senses electrical impulses within the brain through electrodes placed on the scalp and records them. It is a recording of brain activity, which is the result of the activity of billions of neurons in the brain. EEG can help diagnose conditions such as seizure disorders, strokes, brain tumors, head trauma, and other physiological problems. The pattern of EEG activity changes with the level of a person's arousal. A relaxed person has many slow EEG waves whereas an excited person has many fast waves. A standardized system of electrode placement is the international 10-20 system. A common problem with EEG data is contamination from muscle activity on the scalp. It is desirable to remove such artifacts to get a better picture of the internal workings of the brain. In latest publication[20] a new method is proposed for EEG signal classification in BCI systems by using nonlinear ICA algorithm. An ICA based EEG feature extraction and modeling approach for person authentication is presented in [21]. Efficient use of modern DSP and Soft computing tools in the area of medical diagnosis is recently covered in [22].

In this paper we will focus on extracting and removing artifacts from recorded EEG data using Independent Component Analysis.

2. EEG Database and Preprocessing The BIOMED team and department of Clinical and Experimental Neurology, both at Katholieke Universiteit Leuven (KUL) (Belgium) have given public access to two of their long-term EEG recordings from patients suffering from Mesial Temporal Lobe Epilepsy. Patient was 35 years-old male. The data was collected from 21 scalp electrodes placed according to the international 10-20 System with addition electrodes T1 and T2 on the temporal region. The sampling frequency was 250 Hz and an average reference montage was used. The electrocardiogram (ECG) for each patient was also simultaneously acquired and is available in channel 22 of each recording Under this system, the EEG electrodes are placed on the scalp at 10 and 20 percent of a measured distance. For example, if a circumference measurement around the skull was approximately 55 cm, a base length of 10% or 5.5 cm and 20% or 11.0 cm would be used to determine electrode locations around the skull. The skull may be different from patient to patient but the percentage relationships remain the same. Figure 1 shows a typical 10–20 electrode placement looking down on the skull. Each site has a letter and a number or another letter to identify the hemisphere location.The letters Fp, F, T, C, P, and O stand for Front polar, Frontal, Temporal, Central, Parietal and Occipital respectively. Even numbers (2, 4, 6, 8) refer to the right hemisphere whereas odd numbers (1, 3, 5, 7) refer to the left hemisphere. The z refers to an electrode placed on the midline. The smaller the number, the closer the position is to the midline.


97

Figure 1. Typical electrode placements under the International 10 –20 system

The original EEG in f1.EDF from second 7500 to second 8100 (f1_750to810.set, 12 Mbytes). This EEG frame contains a seizure[18][19].

3. Independent Component Analysis(ICA) Independent Component Analysis (ICA) is one of a group of algorithms to achieve blind separation of sources [Jutten & Herault 1991]. ICA finds an unmixing matrix which linearly decomposes the multichannel EEG data into a sum of maximally temporally independent and spatially fixed components. These Independent Components (ICs) account for artifacts, stimulus and response locked events and spontaneous EEG activity. One of the standard applications of ICA to EEG includes artifact detection and removal . Selected components responsible for artifacts are set to zero and all other ICs can be projected back onto the scalp yielding EEG in true polarity and amplitudes. Related approaches to magneto encephalographic signals can be found . Some simple neural network algorithms cane blindly separate mixtures, of independent sources. On maximizing the joint entropy(y), of the output of neural processor minimizes the mutual information among the output components, yi = g(ui), where g(ui) is an invertible bounded nonlinearity and u=Wx, a version of the original sources. ICA is suitable for performing blind source separation on EEG data because: (1) it is possible that EEG data recorded at multiple scalp sensors are linear sums of temporally independent components arising from spatially fixed, distinct brain or extra-brain networks, and, (2) EEG data by volume conduction does not involve significant time delays. In EEG analysis, the rows of the input matrix x are the EEG signals recorded at different electrodes, while the columns are measurements recorded at different time points. 3.1 Types of artifacts Severe contamination of EEG activity by artifacts such as eye movements, blinks, head movements, muscle, and line noise create a problem for proper EEG interpretation and analysis. The three types of muscle artifacts studied in this paper are: 1) Eye artifacts – project mainly to the frontal side 2) Rear head artifacts – project mainly to the occipital Side

3) Muscle artifacts – dispersed throughout the brain .

3.2 Assumptions for the ICA model

The following assumptions ensure that the ICA model estimates the independent components meaningfully. Actually the first assumption is the only true requirement which ICA demands. The other assumptions ensure that the estimated independent components are unique. (1) The latent variables (or independent components) are statistically independent and the mixing is linear. (2) There is no more than one gaussian signal among the latent variables and the latent variables have cumulative density function not much different from a logistic sigmoid . (3) The number of observed signals, m, is greater than or equal to the number of latent variables, n (i.e. m _ n). If n > m, we come to a special category of Independent Component Analysis called ICA with over-complete bases . In such a case the mixed signals do not have enough information to separate the independent components. There have been attempts to solve this particular problem but no rigorous proofs exist as of yet . If m > n then there is redundancy in the mixed signals. The ICA model works ideally when n = m. (4) The mixing matrix is of full column rank, which means that the rows of the mixing matrix are linearly independent. If the mixing matrix is not of full rank then the mixed signals will be linear multiples of one another. (5) The propagation delay of the mixing medium is negligible. 3.3 The ICA model applied to EEG Data In case of EEG signals we have m-scalp electrodes picking up correlated brain signals where we would like to know what effectively independent brain sources produced these signals. The ICA model appears well suited for this scenario because it satisfies most of the model assumptions considered in section 4. Start with assuming that EEG data can be modeled as a collection of statistically independent brain signals. Assumption (5) is valid since volume conduction in the brain is effectively instantaneous and assumption (2) is plausible . In this paper, it will attempt to separate the m-observed EEG signals into n-statistically independent components (thus satisfying assumption (3) and (4)). However, it is questionable to assume whether EEG data recorded from m-electrodes is made up of exactly n-statistically independent components since it ultimately cannot know the exact number of independent components embedded in the EEG data. Nonetheless, this assumption is usually enough to identify and separate artifacts that are concentrated in certain areas of the brain such as eye, temporal, and occipital artifacts . The ICA model tends to have a more difficult time in separating artifacts that are more spaced out over the scalp such as muscle artifacts.


98

While steps< Maxsteps

For i=1:EndofBlock

u=weightsxdata+w0y=1/(1+exp(-u))

weights=weights+Irate*(I-2yu)*weights

wts_blowup=1noChange=1

Weights=>

maxWts?

wts_blowup?

End for loop?

~wts_blowup?

step=0;change=nochange;wts_blowup=0

block=1;Irate=Irate*lowerIrate%restart with

lower Irateweights=identity matrix;

old wts=weights;

oldwtchange=weights-old_wtsstep++

angledelta=0delta=oldwtchange

change=oldwtchange2

1

2

Yes

No

No

YesYes

Yes

Yes

No

Yes

3


99

Irate=nochange?

rnk=rank(data)

display("matrix may not beinvariable")

rank<chans?

Display('data has rank<rnk>,Channelsare not independent')

Display("lower learning rate to <Irate> and starting again")

Step>2 ?

display('setp<step>-Irate<Irate>,wchange<change>')

oldwts=weights;

angledelta>60?

step==1?

step>2 &change<nochange?

change>maxWts?

RETURN

No

Yes

Yes

No

Yesangledelta=cos-1

No

nochangechangeolddeltadelta*

*

Irate=IratexlowerIrate;olddelta=delta;

oldchange=change;Yes

No

olddelta=delta;oldchange=chnage;

Yes

No

laststep=step;step=maxsteps;

Irate=Irate*lowerrate2

Yes

Yes

No

Endwhile?

Yes W=weights RETURN

3 2

1

Figure 2. The Bell & Sejnowski Infomax Algorithm

flowchart. 3.4 THE ICA ALGORITHM Now here we present a brief derivation of the Bell Sejnowski Information Maximization algorithm. It has been consider a simple case of a one-input one-output system to derive the ICA algorithm. The general multi-input multi-output system is similarly derived with n-dimensional matrices of vector-

valued random variables in place of the scalar valued functions. Consider a scalar-valued function x with a gaussian fx(x) that passes through a transformation function y = g(x) to produce the output with fy(y) (Figure 3). This is analogous to our matrix operation:


100

Y = WX For our work with EEG data we will take the transformation function y to be the logistic sigmoid function defined as:

uexgy −+

==1

1)(

0wwxu += where w = slope of y (also called the weight) w0 = bias weight to align the high density parts of the input with y (Refer Figure 2.)

Figure 3. Transformation of the fx(x), of x when x is mixed with a sigmoid mixing function An increase in the joint entropy of the output, H(y),means a decrease in its mutual information. The entropy of the output is maximized when we align the high density parts of pdf of x with the high sloping parts of the function g(x) (hence the need for the biasing weight w0). The function g(x) is monotonically increasing (i.e. has a unique inverse) and thus the output fy(y) can be written as a function of the input fx(x) as:

xy

xfyf x

y

∂∂

=)(

)(

(1) The entropy of the output is given by,

dyyfyfyfEyH yyy )(ln)()}({ln)( ∫∞

∞−

−=−= (2)

Substituting (1) into (2) gives,

)}({lnln)( xfExyEyH x−

∂∂

= (3)

We now would like to maximize H(y) of eq.(3) for statistical independence. Looking at the right hand side we see that the function x is fixed and the only variable we can change is y. Or more preciously, the slope, w, of y. Hence we take the

partial of H(y) with respect to w. The second term in eq.(3) does not depend on w and therefore can be ignored. The change in slope, ∆ w, necessary for maximum change in entropy is then:

∂∂

∂∂

=∂

∂∆

xyE

wwyHw ln)(

α (4)

We now come to an important step. We would like to compute the derivative, but we cannot compute the expectation. Hence, we make the stochastic gradient approximation:

xy

xyE

∂∂

≈

∂∂ lnln

to get rid of the expectation [4],The equation then simplifies to:

∂∂

∂∂

∂∂

=

∂∂∂

=∂

∂∆

−

xy

wxy

xy

dwwyHw

1

ln)(α (5)

The above equation is the general form of the weight change rule for any transformation function y. For the logistic sigmoid function eq.(1), the terms in eq .(5) are evaluated as:

)1( ywyxy

−=∂∂

(6)

))21(1)(1( ywxyyxy

w−+−=

∂∂

∂∂

(7)

Substituting the above equations into eq.(5) gives the weight update rule for y = logistic sigmoid function:

xyww )21(1 −+∆ −α (8) Similarly, the bias weight update, w0, can be evaluated as:

yw 210 −∆ α (9) Following similar steps we can derive the learning rules for multivariate data for a sigmoid function:

[ ] TT xyWW )21(1−+∆

−α (10)


101

yw 210 −∆ α (11)

4. MatLab Implementation Equations (10) and (11) give the learning rules for updating the weights to perform ICA. Implementing them directly into Matlab will involve performing the inverse function, which is computationally very intensive. We therefore modify eq.(10) by multiplying it by WTW (this does not change anything since W is orthogonal):

[ ]

,))21(()21((

))21((

)(

1

WuyIWWWxyIW

WWxyWW

WWW

yHW

T

TT

TTT

T

−+∆⇒

−+∆⇒

−−∆⇒

∂∂

∆

−

α

α

α

α

(12)

Where u=xW The bias weight update rule remains the same:

)21(0 yw −∆ α (13) The proportionality constant in eq.(12) and (13) is called the learning rate (lrate). In summary, the following two weight update rules are used to perform ICA in Matlab:

[ ]WuyIIrateWW Toldnew ))21(( −++=

(14)

)21(00 yIrateWW oldnew −+= (15) Where Irate= Learning Rate ; W=weight matrix ; W0=bias weight ; I=Identity matrix; y=logistic sigmoid ; u=W x data+w0 ;

5. Result Discussion Data Set f1.set considered in the paper contains 600 seconds of data with sampling frequency Fs=250 Hz. There are 21 channels of data. The data was collected from electrodes placed on the scalp at standard locations using the international 10-20 system[4]. The EEG data is plotted using function implemented in Matlab[18] and is depicted in figure(1). This data contains a seizure onset around 300 onwards on T3-T5 channel with the appearance of rhythmic waves. Occipital artifacts on O1and O2 .Eye blink artifact are on Fp1 and Fp2 and Muscle Artifact are on all channels.

Figure 4. EEG data from Data Set (f1.set)

5.1 Independent Components: Execution of the data is processed through Matlab function of EEG toolbox[19]. The resulting independent components are shown in figure(5).

Figure 5. Independent Components of Dataset(f1.set) 5.2 Topographical Projections The topographical projections of independent components are shown in figure (6).

Figure 6. Independent Components with their respective topographical projection of Dataset(f1.set)


102

5.3 Corrected EEG Data Selected the Right ,Left and Frontal Artifact were removed from EEG data using ICA technique. The resulting artifact corrected EEG data is shown in figure 7. A Comparison with the original EEG data (figure 4) clearly shows that the identified muscle artifacts have been greatly reduced

Figure 7. Corrected EEG Data of Dataset (f1.set)

6. Conclusions It is clear that Independent Component Analysis is well suited to perform artifact correction on EEG data. The topographical views provided the first clues as to which components might be artifacts. These plots together with the time plots of the independent components were used to identify the eye and occipital artifacts. One of the unique properties of ICA is that it can eliminate the artifacts alone without disturbing the surrounding EEG activity. An alternate approach for artifact extraction could be simply subtracting the frontal, temporal, and occipital readings from the EEG data. But this would lead to considerable loss in collected information. The muscle artifacts appearing on all channels for Dataset (f1.set) after a seizure onset could not be removed or reduced significantly. One reason could be that these artifacts are not concentrated in any one region alone and hence ICA algorithm cannot interpolated them as originating from any single electrode. ICA technique will be useful in various BCI application such as mental task detection, detecting various brain disorder such as epilepsy and Image preprocessing etc . References [1] M. Ungureanu, C. Bigan, R. Strungaru, V. Lazarescu,

Independent Component Analysis Applied in Biomedical Signal Processing, Measurement Science Review, Vol. 4, Section 2, 2004.

[2] Arnaud Delorme, Scott Makeig, EEGLAB: An open source toolbox for analysis of single-trial EEG dynamics including independent component analysis, Journal of Neuroscience Methods, 134, pp. 9-21, 2004.

[3] In J. Laidlaw, A. Richens, and D. Chadwick, editors, ATextbook of Epilepsy, pages 1–22. Churchill Livingstone, London, 4th edition, 1993.

[4] J. Gotman. Seizure recognition and analysis. In J. Gotman, J.R. Ives, and P. Gloor, editors, Long-term

Monitoring in Epilepsy,volume 37. Elsevier, Amsterdam, eeg supplement no. 37 edition, 1985.

[5] H. G. Wieser. Monitoring of seizures. In M. R. Timberland and E. H. Reynolds, editors, What is Epilepsy? The Clinical and Scientific Basis of Epilepsy, pages 67–81. Churchill Livingstone, London, 1986.

[6] Anthony J Bell & Terrance J Sejnowski, An

information-Maximisation approach to blind separation and blind deconvolution, Neural Computation, 7,6, 1004-1034 (1995).

[7] Dominic Chan, Blind Signal Separation, PhD Dissertation, University of Cambridge Jean-Francois Cardoso Blind Signal Separation: Statistical principles Proceedings of the IEEE, vol 9, no 10. p 2009-2025 Oct 1998.

[8] Scott Makeig, Independent Component analysis of Electroencephalographic Data Advances in Neural Information Processing Systems 8’’. MIT Press, Cambridge MA 1996 .

[9] Foldvary N, Klem G, Hammel J, Bingaman W, Najm I, Lüders H.(2001) The localizing value of ictal EEG in focal epilepsy. HXURORJ\ 57:2022-2028.

[10] Klass DW.(1995) The continuing challenge of artifacts in the EEG. 35: 239–269.

[11] B. Boashash, M. Mesbah, and P. Colditz, "Time Frequency detection of EEG abnormalities," chapter 15, Article 15.5. pp. 663-669, Elsevier 2003.

[12] R.G. Andrzejak, G.Widman, K. Lehnertz, C. Rieke, P. David, C.E. Elger, "The epileptic process as nonlinear deterministic dynamics in a stochastic environment: an evaluation on mesial temporal lobe epilepsy," Epilepsy Res. Vol. 44, pp. 129-140, 2001.

[13] H.D.I. Abarbanel, R. Brown, and M.B. Kennel, "Lyapunov exponents in chaotic systems: Their importance and their evaluation using bserved data," International Journal of Modern Physics, vol. 5(9), pp.1347-1375, 1991

[14] A. Wolf, J.B. Swift, H.L. Swinney, and J.A. Vastano, "Determining Lyapunov exponents from a time series," Physica D, vol. 16(3), pp. 285–317, 1985.

[15] P. Grassberger, T. Schrieber, "Nonlinear time sequence analysis," Int. J. Bifurcat. Chaos 1 (3), pp.512-547, 1991.

[16] N. Kannathala,b, M. L. Choob, U.R. Acharyab, P.K. Sadasivana "Entropies for detection of epilepsy in EEG," Computer Methods and Programs in Biomedicine 2005.

[17] A. Subasi, "Epileptic seizure detection using dynamic Wavelet network," Expert Systems with Applications, vol. 29, pp 343–355, 2005

[18] De Clercq,W,Vergult, A.,Vanrumste B.,Van Paesschen,W.,and Van Huffel, S.’Canonical Correlation analysis applied to remove muscle artifacts from the electroencephalogram’, IEEE T.Biomed.Eng.2006;53:2583-2587.

[19] ergult,A,Delercq,Q.,Palmini,A.,Vanrumaste,B.,Dupont P.,Van Huffel,S.,Van Paesschen,W.’Improving the Interpretation of Ictal Scalp EEG:BSS-CCA Algorithm for Muscle Artifact Removal, Epilepsia 2007;48(5):950-958.


103

[20] Farid Oveisi,”EEG signal classification using nonlinear independent component analysis”,icassp,pp.361-364,2009 IEEE International Conference on Acoustics, Speech and Signal Processing,2009.

[21] Chen He Wang ,J.”An Independent Component Analysis(ICA) Based Approach for EEG Person Authentication” , Proc. Of Bioinformatics and Biomedical Engineering ,IEEE explore , pp 1-2,2009.

[22] Prabhakar Khandait,Narendra Bawane,Shyam Limaye, “Efficient ECG Signal Analysis using Wavelet Technique for Arrhythmia Detechtion: An ANFIS Approach”, Proc. Of SPIE, Vol 7546,pp 75461G1-G6,2010.

Author Profile Ashish B. Sasankar received the MCA and M.phil degrees in Computer Science from HVPM Amravati University in 1996 and 2008,respectively. He is currently pursuing M.Tech in Computer Science from GHRCE,RTM Nagpur University (India).


104

Analysis of Three Phase Four Wire Inverter for UPS Fed Unbalanced Star Connected Load

R.Senthil Kumar1, Dr. Jovitha Jerome2 and S.NithyaBhama3

1 Department of Electrical and Electronics Engineering, Bannari Amman Institute of Technology Anna University, Tamil Nadu India

[email protected]

2 Department of Control and Instrumentation Engineering, PSG College of Technology, Anna University, Tamil Nadu India

[email protected]

3 Department of Electrical and Electronics Engineering, Bannari Amman Institute of Technology Anna University, Tamil Nadu India

[email protected] Abstract: A Three phase inverter with a neutral connection i.e., three phase four wire inverter is proposed. The uninterruptible power supply (UPS) system is fed by three phase four wire inverter and the load neutral point voltage is low to meet the requirement of the system. The four leg inverters effectively provide the neutral connection in three phase four wire system. They are used in many applications to handle the neutral current caused by the unbalanced and non-linear load. The unbalanced load becomes non-linear, where the neutral of the loads are accessible. The four leg inverter produces the three output voltages independently with one additional leg. The main feature of a three phase inverter, with an additional neutral leg, is its ability to deal with load unbalance in a system. The goal of the three phase four leg inverter is to maintain the desired sinusoidal output voltage waveform for all loading conditions and transients. The neutral connection is present to handle the ground current due to unbalanced loads. The feasibility of the proposed modulation technique is verified by MATLAB/SIMULINK.

Keywords: Four wire inverter, Rectifier, THD, UPS.

1. Introduction The primary function of an UPS is to maintain a constant voltage and constant frequency supply for critical loads, irrespective of variations in the input source or load condition [2]. The way of providing a neutral connection for three phase four wire systems using a four leg inverter topology by tying the neutral point to the mid point of the fourth neutral leg. The three phase four inverter has more control flexibility, because two additional power switches doubles the number of inverter output states from 8(=2 3 ) to 16(=2 4 ).This allows to improve the output waveform quality.

In the medium or low power UPS; an output transformer is used to mitigate the neutral to earth voltage. In the high power UPS; it is to eliminate the output transformer so that load is fed by the inverter directly, so the neutral of earth voltage is emerged. The currents flowing on each phase are generally not balanced so, that a transformer is not required,

a connection to the neutral terminal should be provided by adding an extra wire to the inverter.

The load neutral terminal can be connected to the inverter using two topologies: • Three phase four-wire, in which the neutral point is

connected directly to the midpoint of the supply by means of a capacitor divider.

• Three phase four-leg, employing an additional inverter leg that permits to modify the neutral point voltage.

The first topology is certainly simplest one, but the three-phase inverter turns into three independent single-phase inverters. As consequence, zero-sequence harmonics are generated; moreover, especially when the load is unbalanced or non-linear, a high voltage ripple over supply capacitors is produced by neutral currents. A further limitation is represented by the maximum voltage value that the amplitude of each phase fundamental harmonic can reach.

The second topology requires additional power switches and a more complex control strategy, but it offers different advantages, such as an increased maximum output voltage value, a reduction of neutral currents and the possibility of neutral point voltage control [5-7].

The block diagram for the four wire inverter for online UPS as shown in Figure 1.

Figure 1.Block diagram for four wire inverter

The main components of the UPS are rectifier, battery, four wire inverter, four wire inverter and load. When the main supply is present, the rectifier provides power to an inverter as well as battery. The battery is charged. The inverter is on and feeds power to the load through UPS switch. The UPS


105

switch is always on and connects load to inverter output. When the UPS fails, then load is connected directly to the mains directly through main switch. When the supply is not available, then battery bank supplies power to an inverter. Thus an inverter is always on and it takes power from rectifier or battery.

The three phase four wire inverter is suitable for use in high power UPS for its advantage of feeding unbalanced load and the higher dc voltage utilization [3]. As the load is fed by three phase three wire inverter is shown in Figure 2. In this paper, the load neutral point voltage for the three phase four leg inverter is proposed and it is shown in Figure 3.

Figure 2.Three phase three wire inverter

Figure 3.Three phase four wire inverter

2. Three Phase Four Wire Inverter The three phase four wire inverter obtained by replacing the three wire switching network with a four wire switching network is shown in Figure 4.

Figure 4.Four wire Switching Network

The simplified diagram of four leg inverter circuit feeding four wire load is shown in Figure 5.The neutral inductor Ln can reduce switching frequency ripple.

Figure 5.Simplified diagram for four wire inverter

The switch in the inverter legs R,Y,B,N denoted as Sk(SR,SY,SB,SN) corresponds to each vector Vk,for S=1 upper switch in the inverter wire is conducting and for S=0,the lower switch is conducting. The vector V (1011) represents switching state is shown in Figure 5.[8]. The equivalent circuits for states (1011) and (1010) are represented in Figure 6(a) and Figure 6 (b) respectively.

Figure 6(a).For switching state SRSYSBSN 1011

VRN =VBN=0 and VYN=-2Vd

Figure 6(b).For switching state SRSY SBSN 1010 VRN =VBN=2Vd and VYN=0 The comparison of a three phase 3 wire and 4 wire voltage source inverter as shown in table 1.

Table 1: Comparison of three phase 3 wire and 4 wire inverter

S.NO

PARAMETER

THREE PHASE THREE WIRE

LOAD

THREE PHASE

FOUR WIRE LOAD

1.

Number of

required power

switches

6

8

2.

Equivalent

topology

Three

independent

single phase half

bridge.

Three

dependent

single phase full

bridge.

3.

Number of the

output vectors

6(no zero vectors)

16(14 active + 2

zero vectors)


106

4.

Maximum

achievable peak

value of line to

neutral voltage.

0.5Vd

0.577Vd

There are 16 switching states which are listed in table 2; it can be taken from the graphical representation of switching vectors in Figure 7.

There are 14 non-zero voltage vectors and two zero vectors(1111), (0000).The three phase variables Kr,Ky and Kb can be transferred as orthogonal coordinates kα,kβ,kγ using eq (1). Any three phase sinusoidal set of quantities can be transformed to an orthogonal reference.

For given switching states of the inverter, the voltage vector components can be calculated as,

( ) ( )( ) ( )

−−−−

=

b

y

r

kkk

k

kk

21212134sin32sinsin34cos32coscos

32 πθπθθπθπθθ

γ

β

α (1)

Where θ is the angle of orthogonal set α-β-0 with respect to arbitrary reference. If α-β-0 axes are stationary and the α-axis is aligned with the- axis, then θ=0 at all times. Thus, we get

−

−−=

b

y

r

kkk

kkk

21212123230

2121132

γ

β

α (2)

The above matrix can be rewritten as ( )BYRd SSSVV −−= 2.31α

(3)

( )BYd SSVV −= .31β (4)

( )( )BYRNd SSSSVV ++−−= 3.31γ (5)

Table 2: Switching combination and output voltages for

3 phase 4-wire inverter

NO.

SR,SY,SB,SN

Vα

Vβ

Vγ 0

0000

0

0

0

1

0001

0

0

-Vd

2

0010

-1/3Vd

-1/ 3 Vd

1/3Vd

3

0011

-1/3Vd

-1/ 3 Vd

-2/3 Vd

4

0100

-1/3Vd

1/ 3 Vd

-1/3Vd

5

0101

-1/3Vd

1/ 3 Vd

-2/3 Vd

6

0110

-2/3 Vd

0

2/3 Vd

7

0111

-2/3 Vd

0

-1/3Vd

8

1000

2/3 Vd

0

1/3Vd

9

1001

2/3 Vd

0

-2/3 Vd

10

1010

1/3Vd

-1/ 3 Vd

2/3 Vd

11

1011

1/3Vd

-1/ 3 Vd

-1/3Vd

12

1100

1/3Vd

1/ 3 Vd

2/3 Vd

13

1101

1/3Vd

1/ 3 Vd

-1/3Vd

14

1110

0

0

Vd

15

1111

0

0

0

Figure 7.Switching vectors for three phase four wire inverter

3. Circuit Description of Four Wire voltage Source Inverter

The three phase four wire voltage source inverter, commonly used for three phase voltage generation is shown in Figure 8. It consists of eight switches Srp-Sxn and filter of inductor LR-LX and capacitors CR-CB.The LC filter filters out the switching harmonics. The voltage source inverter able to generate balanced and high quality AC output voltage, shown in Figure 8.

Figure 8.Three phase output voltages

In the three phase output voltage waveform shown in Figure 8, one line cycle is divided into six regions. In region 0˚–60˚, 120˚–180˚ and 240˚–300˚, the voltage waveforms in Figure


107

8 have similar pattern, i.e., one-phase voltage is always lower than the other two [2].

The modulation method for four wire inverter are 1) The switch Sin (i = r, y, b) for the phase with the lowest

voltage is always turned ON and the corresponding Sip for this phase is always turned OFF.

2) The switches Sin and Sip for the other two phases are driven complementarily.

3) The switches Sxn and Sxp for the neutral phase are driven complementarily.

The main circuit diagram in Figure 3 is equivalent to Figure 9(a) in 0˚–60˚region, which can be further organized into Figure 9(b). The same equivalent circuit is also applicable to 120˚–180˚ and 240˚–300˚ regions. The switching of the inverter is shown in table 3.

(a)

(b)

Figure 9.Equivalent circuit for four wired VSI for 0˚–60˚

In region 60˚–120˚, 180˚–240˚and 300˚–360˚, the voltage waveforms in Figure 8 have another pattern, i.e., one phase voltage is always higher than the other two [2]. The modulation method for four wire inverter are 1) The switch Sip (i = r, y, b) for the phase with the

highest voltage is always turned ON and the corresponding Sin for this phase is always turned OFF.

2) The switches Sip and Sin for the other two phases are driven complementarily.

3) The switches Sxn and Sxp for the neutral phase are driven complementarily.

With this Figure 3 is equivalent to Figure 10 (a) in 60˚–120˚ region, which can be further organized into Figure 10(b).The same equivalent circuit is also applicable to 180˚–240˚and 300˚–360˚ regions. The switching of the inverter is shown in table 3.

(a)

(b)

Figure 10.Equivalent circuit for four wired VSI for 60˚–120˚

For further analysis, following assumptions are made.

1) LR = LY = LB = LX = L. 2) CR = CY = CB = C. 3) Switching frequency is much higher than fundamental

frequency.

Table 3: Switching logics for proposed controller

SWITCHES

DEGREES

S1

S2

S3

S4

S5

S6

N1

N2

0˚- 60˚

ON

OFF

OFF

OFF

OFF

ON

ON

OFF

60˚-120˚

ON

ON

OFF

OFF

OFF

OFF

OFF

ON

120˚-180˚

OFF

ON

ON

OFF

OFF

OFF

ON

OFF

180˚-240˚

OFF

OFF

ON

ON

OFF

OFF

OFF

ON

240˚-300˚

OFF

OFF

OFF

ON

ON

OFF

ON

OFF

300˚-360˚

OFF

OFF

OFF

OFF

ON

ON

OFF

ON


108

4. Simulation Results The Figure 11 shows the three phase AC rectifier and its output.

Figure 11.Simulation circuit for rectifier.

Figure 12.Simulation result for rectifier

The above rectified output voltage in Figure 12 obtained across the capacitor.

The Figure 13 is the three phase four wire inverter for online UPS is proposed.

Figure 13.Simulation circuit for three phase four wire

inverter

From the simulation analysis of Figure 13 (i) The wire N provides a lower impedance loop for

unbalanced current and triplen harmonics, so the imbalance of output is dramatically reduced.

(ii) The neutral inductance Ln can reduce the current that flows through the Switching components of wire N.

Figure 14.Input voltage for three phase AC source

The Figure 14 is the three phase input source voltage for the UPS.

Figure 15.Simulation result for four wire inverter

The Figure 15 is the simulation result for four wire inverter for three phase each output is phase shifted by 120˚.

The Figure 16 shows the DC source input voltage for four wire inverter.

Figure 16.Simulation circuit for DC source four wire

inverter

Three line voltages VRY, VYB and VBR are step waves, with step height Vdc/2 and Vdc. The three line voltages are mutually phase shifted by 120˚ as shown in Figure 17.


109

Figure 17.Simulation result for three phase four wire

inverter

Table 4: Simulation result parameters

Parameter

Values

Voltage for each phase

100V

Frequency

50HZ

DC input voltage

200V

Inductance(L)

1mH

Capacitance(C)

1000µF

Neutral inductance(Ln)

1mH

Rated resistive load

100Ω

The neutral voltage waveform for four wire inverter as shown in Figure 18.

Figure 18.Simulation result for neutral voltage

The Figure 19 shows the THD level for three Phase four wire system. The harmonic distortion is reduced and its THD level is 3.92%.

Figure 19.THD level for three phase four wire inverter

5. Conclusion The three phase four wire UPS has been proposed in this paper. The fourth wire makes the inverter have the ability of handling unbalancing loads. The inductor in fourth wire reduces the current through the switching components. The inverter control has the advantages of both lower switching to fundamental frequency ratio and outstanding ability to carry unbalanced loads.

References [1] Fanghua Zhang, and Yangguang Yan “Selective

Harmonic Elimination PWM Control Scheme on a Three-Phase Four-Leg Voltage Source Inverter” IEEE Trans. Power Electronics, vol. 24, no. 7, July 2009.

[2] Lihua Li and Keyue Ma Smedley, “A New Analog Controller for Three-Phase Four-Wire Voltage Generation Inverters” IEEE Trans. Power Electronics, vol. 24, no. 7, July 2009.

[3] Liu Zeng, Liu Jinjun and Li Jin “Modeling, Analysis and Mitigation of Load Neutral Point Voltage for Three-phase Four-leg Inverter” IPEMC2009.

[4] Salvador Ceballos, Josep Pou, Jordi Zaragoza, José L.

Martín, Eider Robles, Igor Gabiola, and Pedro Ibanez, “Efficient Modulation Technique for a Four-Leg Fault-Tolerant Neutral-Point-Clamped Inverter” IEEE Trans. Industrial Electronics, vol. 55, no. 3, March 2008.

[5] Armando Bellini and Stefano Bifaretti “Modulation Techniques for Three-Phase Four-Leg Inverters” Proceedings of the 6th WSEAS International Conference on Power Systems, Lisbon, Portugal, September 22-24, 2006.

[6] Bellini and S. Bifaretti “A Simple Control Technique for three phase four leg inverters”. SPEEDAM 2006.

[7] Richard Zhang, V. Himamshu Prasad, Dushan Boroyevich and Fred C.Le “Three-Dimensional space Vector Modulation for Four –Leg Voltage-Source Converters” IEEE Trans.Power Electronics, vol.17, no.3, May 2002.

[8] Salem M. Ali Marian and P. Kazmierkowski “PWM Voltage and Current Control of Four-Leg VSI” 1998 IEEE.


110

Authors Profile

SenthilKumar.R was born in Tamilnadu, India, on November 2, 1966. He received the B.E degree in Electrical and Electronics Engineering from Madurai Kamaraj University, in 1989. He received his M.E (Power systems) from Annamalai University, in 1991. He has 15 yrs of teaching experience. Currently he is working as Asst. Professor in EEE department, Bannari

Amman Institute of Technology Sathyamanglam. Currently he is doing research in the field of power converters for UPS Applications. Dr.Jovitha Jerome was born in Tamilnadu, India, on June 2, 1957. She received the B.E. degree in Electrical and Electronics Engineering and M.E. degree in Power Systems from College of Engineering, Guindy, Chennai. She did her DEng in Power Systems. Presently she is working as Professor

and Head in Instrumentation and Control Engineering Department of PSG College of Technology, Coimbatore.

NithyaBhama.S was born on September 4, 1987. She received her B.E Degree in Electrical and Electronics Engineering from Erode Sengunthar Engineering College, Thudupathi, Anna University. Currently she is pursuing M.E in Power Electronics and Drives at Bannari Amman Institute of Technology, affiliated to Anna University.


111

Self Encrypting Data Streams for Digital Signals

A.Chandra Sekhar* Ch.Suneetha1 G.Naga Lakshmi2

*Professor,Department of Engineering Mathematics, GITAM University,Visakhapatnam,India [email protected]

1,2Asst.Professor,Department of Engineering Mathematics, GITAM University,Visakhapatnam,India

Abstract: Cryptography plays a vital role in implementing electronic security systems. Cryptography can be used for signing electronic documents digitally, digital rights management, banking and for controlling access key documents. Public key cryptography is secure if and only if extracting secret key from public information is intractable. Hence design of easily replicable cryptographic primitives is essential. In this paper a new technique for self encrypting data streams is proposed basing on the matrices and quadratic forms. This algorithm uses a quadric form as private key. The key matrix is also sent along with the cipher text in form of a string, which is further encrypted. The elements of the key matrix will be different for different data streams. This consists of data encryption at two levels and the cipher text so obtained becomes quite impossible to break or to interrupt. Keywords: Quadratic forms, cryptography, key matrix

1. Introduction Public key cryptography was introduced by Diffie and

Helman[3][4] In public key cryptosystems, the decryption keys must be kept secret. A decryption key is therefore called a “secret key” or a “private key”. Some important public key cryptosystems are RSA cryptosystem, Rabin encryption, El Gamal encryption. But the known Public key cryptosystems are not efficient as many symmetric cryptosystems. Therefore in practice, hybrid cryptosystems. i.e., combinations of public key systems and symmetric systems are used. For example, Alice if wants to send a message m in encrypted form to Bob. She generates a session key for an efficient symmetric cryptosystem. Then she encrypts the message m using the session key and the symmetric system obtaining the cipher text C. This encryption is fast because an efficient symmetric cryptosystem has been used. Alice also encrypts the session key with Bob’s public key. Since the session key is small, this encryption is also fast. Bob decrypts the session key using his private key. Then he decrypts the cipher text C with the session key and obtains the original message M. In this paper we applied this method to the matrices and quadratic forms.

2. Quadratic Form A homogeneous Polynomial of the second degree in n variables i.e., the expression

jiij

n

1j

n

1i

xxa∑∑==

=φ

where aij ∈R and aij = aji is called a quadratic form in n variables x1, x2,…,xn over a real field. Thus

n1n121122111 xxa.......xxaxa +++=φ

n2n22

2221221 xxa.......xaxxa ++++ +………………….. +………………….. …………………...

2nnn2n2n1n1n xa.......xxaxxa ++++

In the quadratic form there are n square terms x12, x22……xn2 and xC2 product terms x1x2, x2x3…..,xn-1xn so there are

2)1n(n

2)1n(nnnCn 2

+=

−+=+ terms. For our discussion

we confine to a quadratic forms of three variables. .gzx2fyz2hxy2czbyax 222 +++++=φ

We know very well that a homogenous second degree equation

0gzx2fyz2hxy2czbyax 222 =+++++=φ represents a pair if planes if and only if. abc + 2fgh – af2 –bg2 – ch2 = 0 that is

0cfgfbhgha

= .

From this we can say that any quadratic form can be written in the form of a matrix as X’ A X where

=

=

cfgfbhgha

Azyx

X .

Here A is a symmetric matrix. So, a symmetric matrix can be recognized as representative of a quadratic form. A is know as the discriminate of a quadratic form

If A ≠ 0 then we say the quadratic form is non singular, in this paper we deal with non singular quadratic forms. Let x1,x2-----xn and y1,y2------yn be two sets of variables consider the following system of linear equations. y1 = b11x1 + b12x2 +----- b1nxn y2 = b21x1 + b22x2 +----- b2nxn ------------------------------------------- ------------------------------------------- yn = bn1x1 + bn2x2 +----- bnnxn Expressing each yi in terms of x1, x2 ------xn. This system of linear equations can be written as matrix equations Y = B X where


112

=

n

2

1

yyy

Y

−−−−−−−−−−−−−−−

−−−−−−

=

nn2x1n

n22221

n11211

bbb

bbbbbb

B and

=

n

2

1

xxx

X

If B is non singular, than we can write X = B-1Y. Which shows that these is one–to–one correspondence between the pairs of vectors X and Y. In this context we say that the system Y = BX determines a linear transformation and where B is called the matrix of linear transformation. 2.1 Linear transformation of a Quadratic form Let AX'X=ϕ be a real Quadratic form in n variables. Since A is a symmetric matrix ϕ can be reduced by a non singular linear transformation X = PY over the real field[1][2][3][4]. Now X’ A X = (PY)’A (PY) = Y’(P’ A P)Y = Y’BY [where B = P’ A P] We note that B’= (P’ A P)’ = P’ A (P)’= P’ A P = B Clearly Y’ B Y is also a Quadratic form over the real field. Thus if a Quadratic form AX'X is subjected to linear transformation X = PY then the transformation is also a Quadratic form Y’B Y where B = P’A P

3. Proposed Method

Alice wants to send a message M to Bob. She convents the message with equivalent numbers according to the following table[7][8][9].

She chooses a non singular Quadratic form which is otherwise known as a private key in such a way that, the message matrix M and the corresponding matrix of the Quadratic form Q are conformable for multiplication. She multiplies M with Q resulting a matrix N. Then the elements of N are reduced to mod 27 which gives the encrypted message matrix E. This encrypted matrix E is sent to Bob in public channel. Along with this she also sends a session key matrix K, whose elements are the integer parts of the elements of matrix N with respects to mod27. To maintain absolute secrecy this key matrix K is sent in form of a string which is further reduced to binary numbers. Bob with the help of the cipher text and secret key matrix K and the private key Q and gets the original message back.

4. Algorithm 1 The non singular quadratic form is chosen. 2 The matrix Q of the quadratic is obtained. 3 The plain text is converted into its equivalent

message matrix M which is multiplied with Q to obtain N.

4 The result N is adjusted to mod(27) to obtain E. 5 The key matrix K is obtained by the

adjustment factor for each entry 6 The key matrix K and E are mixed to get the

cipher 7 For decryption , from the cipher K and E are

separated to get N 8 N is multiplied with the inverse of the matrix Q

and adjusted to mod(27) to get the required plain text M.

4.1 Example 4.1.1 Encryption If Alice wants to send a message GOOD LUCK to Bob she converts message to a matrix

M = [6 14 14; 3 26 11; 20 2 10] She takes a non-singular Quadratic form 2x2 + 3y2 + 4z2 + 2xy + 4yz + 4zx Q = [2 1 2;1 3 2;2 2 4] N = M * Q=[54 76 96;54 103 102; 62 46 84] N is reduced to mod 27 resulting E = [0 22 5; 0 22 21; 8 19 3 ] And key matrix K= [2 2 3;2 3 3; 2 1 3] The cipher text C = [0 2 22 2 5 3 0 2 22 3 21 3 8 2 19 1 3 3 ] Which is equivalent to ACWCFDACWDVDICTBCC 4.1.2 Decryption The message ACWCFDACWDVDICTBCC is converted into the equivalent code to obtain the cipher text C =[0 2 22 2 5 3 0 2 22 3 21 3 8 2 19 1 3 3 ] The Key matrix K and E are separated as K=[2 2 3;2 3 3; 2 1 3] E== [0 22 5; 0 22 21; 8 19 3 ]

By multiplying the key with 27 and on adding the cipher we get N = [54 76 96;54 103 102; 62 46 84] By multiplying N with inverse of Q we get the original message as [6 14 14 32 6 11 20 2 10] which is GOOD LUCK.

5. Conclusions

In the new algorithm technique for self encrypting data streams is proposed basing on the matrices and quadratic forms. This algorithm uses a quadric form as private key. The key matrix is also sent along with the cipher text in form of a string, which is further encrypted. This consists of data encryption at two levels and the cipher text so obtained becomes quite impossible to break or to interrupt.Thus the computational overhead is very low. It is almost impossible to extract the original information in the proposed method even if the algorithm is known.

A B C D E F G H I J K L M 0 1 2 3 4 5 6 7 8 9 10 11 12 N O P Q R S T U V W X Y Z 13 14 15 16 17 18 19 2

0 21 22 23 24 25

Null or Space 26


113

References

[1] K.R.Sudha, A.Chandra Sekhar and Prasad Reddy.P.V.G.D “Cryptography protection of digital signals using some Recurrence relations” IJCSNS International Journal of Computer Science and Network Security, VOL.7 No.5, May 2007 pp 203-207

[2] A.P. STAKHOV, ”THE ‘‘GOLDEN’’ MATRICES AND A NEW KIND OF CRYPTOGRAPHY”, CHAOS, SOLTIONS AND FRACTALS 32 ( (2007) PP1138–1146

[3]. A.P. Stakhov. “The golden section and modern harmony mathematics. Applications of Fibonacci numbers,” 7,Kluwer Academic Publishers; (1998). pp393–99.

[4]. A.P. Stakhov. “The golden section in the measurement theory”. Compute Math Appl; 17(1989):pp613–638.

[5]. Whitfield Diffie And Martin E. Hellman, New Directions in Cryptography” IEEE Transactions on Information Theory, Vol. -22, No. 6, November 1976 ,pp 644-654

[6]. Whitfield Diffie and Martin E. Hellman “Privacy and Authentication: An Introduction to Cryptography” PROCEEDINGS OF THE IEEE, VOL. 67, NO. 3, MARCH 1979,pp397-427

[7]. C. E. SHANNON Communication Theory of Secrecy Systems The material in this paper appeared in a confidential report “A Mathematical Theory of Cryptography” dated Sept.1, 1946, which has now been declassified.

[8]. E. Shannon, A Mathematical Theory of Communication, Bell System Technical Journal 27 (1948) 379–423, 623–656.

[9]. A. Chandra Sekhar , ,K.R.Sudha and Prasad Reddy.P.V.G.D “Data Encryption Technique Using Random Number Generator” Granular Computing, 2007. GRC 2007. IEEE International Conference, on 2-4 Nov. 2007 Page(s):573-576

Authors Profile

Dr.A .Chandra Sekhar received his PhD degree in number theory from JNT University and MSc., degree with specialization in algebraic number theory from Andhra University . He Secured the prestigious K.NAGABHUSHANAM Memorial Award in M.Sc., for obtaining University First rank. He

did his MPhil from Andhra University in 2000.He was with Gayatri degree college during 1991to 1995 and later joined GITAM Engineering college in 1995. Presently he is working as Professor and Head of the Department of Engineering Mathematics at GITAM Engineering college, Visakhapatnam, INDIA.

Mrs.Ch.Suneetha is presently working as Assistant Professor in the Department of Engineering Mathematics . She is pursuing her PhD in number theory and cryptography under the guidance of Dr.A.Chandra Sekhar

Mrs.Naga Lakshmi is working as Assistant Professor in the Department of Engineering Mathematics . She is pursuing her MPhil in number theory and cryptography under the guidance of Dr.A.Chandra Sekhar

vol 2 no 4

Documents