security mechanisms for distributed computing systems

43
Security Mechanisms for Distributed Computing Systems A9ID1007, Xu Ling Kobayashi Laboratory GSIS, TOHOKU UNIVERSITY 2011/12/15 1

Upload: medea

Post on 24-Feb-2016

37 views

Category:

Documents


0 download

DESCRIPTION

2011/12/15. Security Mechanisms for Distributed Computing Systems. A9ID1007, Xu Ling Kobayashi Laboratory GSIS, TOHOKU UNIVERSITY. Chapter 1 I ntroduction. Distributed Computing System. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Security Mechanisms for Distributed Computing Systems

1

Security Mechanisms for Distributed Computing Systems

A9ID1007, Xu LingKobayashi Laboratory

GSIS, TOHOKU UNIVERSITY

2011/12/15

Page 2: Security Mechanisms for Distributed Computing Systems

2

Chapter 1

Introduction

Page 3: Security Mechanisms for Distributed Computing Systems

3

Distributed Computing System

• Distributed computing systems (DCSs): A system where nodes cooperate with each other to finish certain goals

• Example: volunteer computing system

Page 4: Security Mechanisms for Distributed Computing Systems

4

Volunteer Computing System

• Utilizes the idling computing resources on the network to finish computing intensive tasks

worker 1 worker 2 worker 3 worker 4

host

Task 1 Task 2 Task 3 Task 4Result 1 Result 2 Result 3 Result 4

The structure of a typical volunteer computing system

Task 1Task 2

Task n

Page 5: Security Mechanisms for Distributed Computing Systems

5

Classification of DCSs• Centralized DCSs:

– Feature: • has some central servers to manage the system• easy to manage

– Example: Volunteer computing system

• Decentralized DCSs: – Features:

• has no central server; all nodes take the same responsibility to maintain the system

• High scalability– Example: Peer to peer system (Skype, bitTorrent)

Page 6: Security Mechanisms for Distributed Computing Systems

6

Security Problems in DCSs• In a DCS, malicious nodes may exist and will launch

attacks against the system

• Two representative attacks against DCSs– False result attack (for both centralized and decentralized DCSs)

• Malicious nodes send incorrect data to other nodes deliberately to interrupt the system

– Sybil attack (mainly in decentralized DCSs)• In decentralized DCSs, it is easy for malicious nodes to join the system• A large number of malicious nodes collude to break the system

protocol• Malicious nodes are called Sybil nodes in this attack

Page 7: Security Mechanisms for Distributed Computing Systems

7

Research Objective

• Address the critical security problems of DCSs– False result attack– Sybil attack

• Approach– For false result attack: enable honest nodes to detect

malicious nodes honest nodes can expel malicious nodes out of the system

– For Sybil attack: enable honest nodes to detect Sybil nodes Honest nodes can expel Sybil nodes out of the system

Page 8: Security Mechanisms for Distributed Computing Systems

8

Dissertation Organization• Chapter 1 Introduction

• Chapter 2 MSC: A False Result Resisting Algorithm

• Chapter 3 SybilDetector: A Shortest Path and Attack Edge Detecting Based Sybil Detecting Algorithm

• Chapter 4 RSSR: A Random Walk and Attack Edge Detecting Based Sybil Detecting Algorithm

• Chapter 5 Conclusion

The Sybil attack

The false result attack

Advantage

SybilDetector Higher efficiency

RSSR No need of trustful third party

Page 9: Security Mechanisms for Distributed Computing Systems

9

Chapter 2

MSC: A False Result Resisting Algorithm

Page 10: Security Mechanisms for Distributed Computing Systems

10

Introduction

• Computing model of DCSs– The system has a host node and multiple worker nodes– The host dispatches tasks to workers– Workers compute the received tasks and return answers to

the host• Problem

– Malicious workers may return incorrect answers to the host– Malicious workers can collude (cooperate to find the best

strategy) to break the system protocol• Objective: enable the host to detect malicious workers

Page 11: Security Mechanisms for Distributed Computing Systems

11

Existing Solution

• Quiz-Based Spot Checking– The host dispatches multiple tasks to each worker v– These tasks contains some special tasks called quizzes– The host checks the correctness of the answers of quizzes

Node v is honest only if the answers of the quizzes return by v are correct

– Problem: • A Quiz should satisfy: the correctness of the answer of a

quiz should be easy to check• How to generate quizzes that satisfy this property is an

open problem.

Page 12: Security Mechanisms for Distributed Computing Systems

12

Address the Problems of Existing Solutions

• Quiz-Based Spot Checking– Dispatches a task set to

each worker v

– The task set contains some special tasks called quizzes

– Checks the correctness of the answers of quizzes

Let workers themselves to check the correctness of the quizzes

Use normal tasks as quizzes

Page 13: Security Mechanisms for Distributed Computing Systems

13

Algorithm OutlineThe host• Dispatches a task set to each

worker. • For each pair of two workers, v and

u, the task sets of v and u have some tasks in common (checking tasks)

• Increases the reliabilities of v and u if v and u return equal answers to their checking tasks (made a match).

Let workers themselves to check the correctness of the quizzes

Use normal tasks as quizzes (checking tasks)

Malicious workers make more mismatches have lower reliabilities be detected

Page 14: Security Mechanisms for Distributed Computing Systems

14

Analysis

• More practical than Quiz-based spot checking– No need to generate the quizzes

• Effective: – Without collusion : Can detect all malicious workers– Under colluding: Can detect all malicious workers when

malicious workers are less than half of the system

Page 15: Security Mechanisms for Distributed Computing Systems

15

Evaluation

• Performance metric– average reliability of honest nodes – average reliability of malicious nodes

• Questions to answers– How will the performance changes as the number

malicious workers increases?

Page 16: Security Mechanisms for Distributed Computing Systems

16

Strategy of Workers

• Honest worker– always return correct answer

• Malicious non-conspirator– compute w% of the tasks it received

• Malicious conspirator– Two conspirators can find and compute their

common checking tasks

Page 17: Security Mechanisms for Distributed Computing Systems

17

Change of Performance as the Number of Malicious Workers Increases

0. 00

0. 20

0. 40

0. 60

0. 80

1. 00

1. 20

0.40 0.45 0.50 0.55 0.60 0.63 0.68 0.73 0.78 0.83 0.88 0.93 0.98

Rel

iabi

lity

Pf

Reliability - Pf (w=0.4, Pc=0.5)

Honest Conspirator Non-Conspirator

• Number of malicious workers is small honest workers have highest reliabilities.

• Number of malicious worker is large conspirators have the highest reliabilities.

Under collusion: MSC can detect malicious nodes when # of malicious nodes is small (50% of the system)

Pf: Percentage of malicious workers in the system

Page 18: Security Mechanisms for Distributed Computing Systems

18

Conclusion

• Objective: design an algorithm to detect malicious workers for the false result attack

• MSC, a malicious worker detecting algorithm– More practical than previous solution– Effective:

• Without collusion : Can detect all malicious workers• Under colluding: Can detect all malicious workers when malicious

workers are less than half of the system

PublicationLing Xu, Hirouyki Takizawa, and Hiroaki Kobayashi: “A Reliability Model for Result Checking in Volunteer Computing”, Proceedings of DAS-P2P 2008 Workshop, pp.201-204, 2008.

Page 19: Security Mechanisms for Distributed Computing Systems

19

Chapter 3

SybilDetector: A Shortest Path and Edge Detecting Based Sybil Detecting Algorithm

Page 20: Security Mechanisms for Distributed Computing Systems

20

Introduction

• Problem: Sybil attack is a critical threat to DCSs– A large number of malicious nodes exist in the system, and

collude to interrupt the system– Collude nodes are called Sybil nodes

• Objective: a Sybil detecting algorithm – Enables each honest node to detect other Sybil nodes

Page 21: Security Mechanisms for Distributed Computing Systems

21

Related Work

• Social network model[Yu2008]

– nodes of the same types are closely connected– nodes of different types are connected by a small number of

attack edges

• Example[Danezis05]: in many P2P systems– To join the system, the new node has to attach to an existing node– Sybil node can only attach to existing Sybil nodesThe number of attack edges is small

Honest cluster Sybil cluster

Attack edges

Page 22: Security Mechanisms for Distributed Computing Systems

22

Assumption• The system obeys the social network model• The system has trustful third parties

Page 23: Security Mechanisms for Distributed Computing Systems

23

SybilDetector

• Observation– For node v, node u is Sybil (v,u)-SP will pass the attack edges

(v,u)-SP: a shortest path between the v and u

• Idea: For v to decide whether u is Sybil– Computes (v,u)-SPs – Detect the attack edges– Judge whether the (v,u)-SPs have passed the attack edges

Honest cluster Sybil cluster

vu

Page 24: Security Mechanisms for Distributed Computing Systems

24

Algorithm: For v to decide whether u is Sybil

• Computes (v,u)-SPs – Use existing distributed

shortest path computing algorithms

• Detect the attack edges– Compute the shortest path betweenness of each edge

SPB of edge e: # of shortest paths that pass e– Attack edges have higher SPBs

• (v,u)-SPs have passed edges of high betweennesses u is Sybil

v uaee

b(ae) = 18

b(e) = 8

sp

Page 25: Security Mechanisms for Distributed Computing Systems

25

Evaluation• Compare the performances of SybilDetector and SybilLimit

• SybilLimit: a representative Sybil detecting algorithm• Performance metric

• Sybil accept rate (sar): the probability that honest node regard Sybil nodes to be honest

• Questions to answer: • What is the influence of the number of attack edges in the

system?• What is the influence of the number of Sybil nodes in the

system?

Honest cluster Sybil cluster

Page 26: Security Mechanisms for Distributed Computing Systems

26

Network Configuration

• Create the honest region: A real world network topology

• Create the Sybil region: synthetic network topologies• Connect the two regions with attack edges

Honest cluster Sybil cluster

Type Node number

Edge number

Real world social network topology

1222 16714

Synthetic random network

500 1725

Honest region

Page 27: Security Mechanisms for Distributed Computing Systems

Change of SAR as the Number of Attack Edges in the System Increases

• SAR increases with g– The btns of attack edges decrease– Less Sybil are detected

• SAR(SybilDetector)<<SAR(SybilLimit)– 50x improvement

10x decrease in SAR

0

0.2

0.4

0.6

0.8

1

1.2

12 36 61 85 109134158183207232256g

real1222rn500, SAR

sar(SybilLimit)

sar(SybilDetector)

50x decrease in SAR

27

Page 28: Security Mechanisms for Distributed Computing Systems

28

Change of SAR as the Number of Sybil Nodes in the System Increases

• As snn increases, SAR of SD decreases– The btns of attack edges increase– More Sybil node detected

• SAR(SybilDetector)<<SAR(SybilLimit)– 4x~180x improvement

0

0.2

0.4

0.6

0.8

1

1.2

snn

real1222g36, SAR

sar(SybilLimit)

sar(SybilDetector)

180 x decreases in SAR

4 x decreases in SAR

Page 29: Security Mechanisms for Distributed Computing Systems

29

Conclusion

• Sybil attack is a critical threat to decentralized DCSs• Objective: enable each honest node to detect Sybil

nodes• Proposed SybilDetector, a Sybil resisting algorithm

– Made great (4x~180x in the simulation) increase in the Sybil detecting accuracy, compared with the representative existing solution

PublicationLing Xu, Satayapiwat Chainan, Hiroyuki Takizawa, Hiroaki Kobayashi, ”Resisting Sybil Attack By Social Network and Network Clustering,” saint, pp.15-21, 2010 10th IEEE/IPSJ International Symposium on Applications and the Internet, 2010

Page 30: Security Mechanisms for Distributed Computing Systems

30

Chapter 4

RSSR: A Random Walk and Attack Edge Detecting Based Sybil Detecting Algorithm

Page 31: Security Mechanisms for Distributed Computing Systems

31

Introduction

• SybilDetector– An effective Sybil detecting algorithm – Problem: needs trustful third parties (to compute

shortest paths)unpractical in some DCSs (e.g., P2P systems)

• Objective: A Sybil detecting algorithm– Remove the need of trustful 3rd partydo not use shortest path use random walk instead

Page 32: Security Mechanisms for Distributed Computing Systems

32

Existing Sybil Detecting Algorithm: SOHL

• Probing random walk: a message packet that moves in a random walk manner for a short distance– Has a low escape rate– Ends at a random honest node w.h.p

• SOHL: each honest node v – Disseminates a large number of probing random walks– Ending nodes of the random walks honest nodes w.h.p– Other nodes Sybil nodes w.h.p

vu Number of attack

edges is smallescape rate is low

Page 33: Security Mechanisms for Distributed Computing Systems

33

Idea

• Improve the performance of SOHL– Detect the attack edges – Prevent probing random walks from passing the

detected edges

vu Number of attack

edge is smallescape rate is low

Page 34: Security Mechanisms for Distributed Computing Systems

34

RSSR• Detect attack edges

– Compute the Random walk betweenness • Each pair of nodes disseminate a random walk

to each other.• RWB of e: the number of times random walk

messages passing e– Attack edges have higher RWB’s

• Prevent probing random walks from passing the detected edges– Reduce the probability that probing

random walks passing the edges of high betweennesses

Do not compute shortest path betweenness

No need of trustful 3rd party any more

Page 35: Security Mechanisms for Distributed Computing Systems

35

Evaluation• Compare the performance of SOHL and RSSR• Performance metric

• Sybil accept rate: the probability that Sybil nodes are regarded honest nodes.

• Questions to answer: • What is the influence of number of attack edges in

the system on the performance?

Honest cluster Sybil cluster

Page 36: Security Mechanisms for Distributed Computing Systems

36

Network Configuration• Create the honest region: A real world network topology• Create the Sybil region: synthetic network topologies• Connect the two regions with attack edges

Honest cluster Sybil cluster

Type Node number

Edge number

Real world social network topology

1222 16714

Synthetic random network

500 1725

Honest region

Page 37: Security Mechanisms for Distributed Computing Systems

37

Change of Sybil Accept Rate as the Number of Attack Edges Increases

• As g increases, SAR increases– Average btns of attack edges decreases– Escape rate increases– Accept more Sybil nodes

• SAR(RSSR) << SAR(SOHL)– Attack edges can be effectively detected

00.10.20.30.40.50.60.70.80.9

1

12 36 61 85 109

134

158

183

207

232

256

281

305

329

354

378

403

427

452

g

real1222rn500, SAR

sar(sohl)

sar(rssr)

3x decreases in SAR28x decrease in SAR

Honest cluster Sybil cluster

Page 38: Security Mechanisms for Distributed Computing Systems

38

Conclusion

• Objective: a Sybil detecting algorithm without trustful 3rd parties

• Proposed RSSR, a Sybil detecting algorithm– Practical: no need of trustful third party

• Shortest path (SybilDetector) Random Walk (RSSR)– Effective: increased the Sybil detecting accuracy of a

representative existing solution by 3~28 timesPublicationLing Xu, Ryusuke EGAWA, Hiroyuki TAKIZAWA, Hiroaki KOBAYASHI, “A Network Clustering Algorithm for Sybil-Attack Resisting”, IEICE Transactions, special section, Parallel and Distributed Computing and Networking

Page 39: Security Mechanisms for Distributed Computing Systems

39

Chapter 5

Conclusion

Page 40: Security Mechanisms for Distributed Computing Systems

40

Conclusion

• Objective: address critical security threats to DCSs– False result attack– Sybil attack

• Approach– Detect malicious workers False result attack

• MSC (Chapter 2)– Detect Sybil nodes Sybil attack

• SybilDetector (Chapter 3)• RSSR (Chapter 4)

Page 41: Security Mechanisms for Distributed Computing Systems

41

Future Work

• Short term – More theoretical analysis of the performances of

SybilDetector and RSSR• Long term

– Extend the social network model (more than two clusters)

Page 42: Security Mechanisms for Distributed Computing Systems

42

Publication• Ling Xu, Hirouyki Takizawa, and Hiroaki Kobayashi: “A Reliability Model for

Result Checking in Volunteer Computing”, Proceedings of DAS-P2P 2008 Workshop, pp.201-204, 2008.

• Ling Xu, Satayapiwat Chainan, Hiroyuki Takizawa, Hiroaki Kobayashi, ”Resisting Sybil Attack By Social Network and Network Clustering,” saint, pp.15-21, 2010 10th IEEE/IPSJ International Symposium on Applications and the Internet, 2010.

• Ling Xu, Ryusuke EGAWA, Hiroyuki TAKIZAWA, Hiroaki KOBAYASHI, “A Network Clustering Algorithm for Sybil-Attack Resisting”, IEICE Transactions, special section, Parallel and Distributed Computing and Networking.

Page 43: Security Mechanisms for Distributed Computing Systems

43

Reference• [Yu2008] Haifeng Yu, Phillip B. Gibbons, Michael Kaminsky,

and Feng Xiao. Sybillimit: A near-optimal social network defense against sybil attacks. In Proc. of IEEE Security & Privacy, 2008

• [Danezis05] G. Danezis, C. Lesniewski-Laas, M. F. Kaashoek, and R. Anderson. Sybil-resistant DHT routing. In European Symposium On Research In Computer Security, 2005