integrated social and quality of service trust management of mobile groups in ad hoc networks...

28
Integrated Social and Quality of Service Trust Management of Mobile Groups in Ad Hoc Networks Ing-Ray Chen, Jia Guo, Fenye Bao, Jin-Hee Cho Communications Surveys & Tutorials, IEEE 13.4 (2011): 562-583 Speaker: Liang Zhao

Upload: eleanore-terry

Post on 25-Dec-2015

213 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Integrated Social and Quality of Service Trust Management of Mobile Groups in Ad Hoc Networks Ing-Ray Chen, Jia Guo, Fenye Bao, Jin-Hee Cho Communications

Integrated Social and Quality of Service Trust Management of Mobile Groups in Ad Hoc Networks

Ing-Ray Chen, Jia Guo, Fenye Bao, Jin-Hee Cho

Communications Surveys & Tutorials,

IEEE 13.4 (2011): 562-583

Speaker: Liang Zhao

Page 2: Integrated Social and Quality of Service Trust Management of Mobile Groups in Ad Hoc Networks Ing-Ray Chen, Jia Guo, Fenye Bao, Jin-Hee Cho Communications

Outlineโ€ข 1.Backgroundโ€ข 2.Trust Management Protocolโ€ข 3.Model-based Evaluation Technique.โ€ข 4.Evaluation Resultsโ€ข 5.Conclusion and Future Works

Page 3: Integrated Social and Quality of Service Trust Management of Mobile Groups in Ad Hoc Networks Ing-Ray Chen, Jia Guo, Fenye Bao, Jin-Hee Cho Communications

BackgroundTrust Management:

A mobile ad hoc network (MANET), sometimes called a mobile mesh network, is a self-configuring network of mobile devices connected by wireless links. 

Mobile Ad Hoc Network (MANETs):

1. Abstract system that processes symbolic representations of social trust2. Aid automated decision-making process.

Page 4: Integrated Social and Quality of Service Trust Management of Mobile Groups in Ad Hoc Networks Ing-Ray Chen, Jia Guo, Fenye Bao, Jin-Hee Cho Communications

Problems in MANET trust Management

โ€ข 1. Traditional QoS Trust Metrics did not consider Social Trust as metric.

โ€ข 2. Existing trust Metrics lack good aggregation parameter settings.

โ€ข 3. Effectiveness of Trust Management Protocol is hard to be evaluated due to difficulty of getting labels based on ground truth.

Page 5: Integrated Social and Quality of Service Trust Management of Mobile Groups in Ad Hoc Networks Ing-Ray Chen, Jia Guo, Fenye Bao, Jin-Hee Cho Communications

Contributionsโ€ข 1. Consider social metrics: i.e. intimacy (social ties)

and honesty (healthiness).โ€ข 2. Identify best trust aggregation parameter settings

for each trust metric.

โ€ข 3. For validating proposed trust management protocol, a novel model-based evaluation technique is leveraged to generate ground truth.

SQTrust

Model-based Evaluation

Page 6: Integrated Social and Quality of Service Trust Management of Mobile Groups in Ad Hoc Networks Ing-Ray Chen, Jia Guo, Fenye Bao, Jin-Hee Cho Communications

SQTrust: A New Trust Management Protocol

Page 7: Integrated Social and Quality of Service Trust Management of Mobile Groups in Ad Hoc Networks Ing-Ray Chen, Jia Guo, Fenye Bao, Jin-Hee Cho Communications

SQTrust: Preliminary

1. Social Ties (Intimacy)

2. Honesty (Healthiness)

measure the social trust level of a node as these social properties are considered critical for trustworthy mission execution

Most important metrics to measure the QoS trust level of a node

3. Competence (Energy)

4. Protocol compliance (Cooperativeness)

Trust Metrics (trust components) taken into account:

Page 8: Integrated Social and Quality of Service Trust Management of Mobile Groups in Ad Hoc Networks Ing-Ray Chen, Jia Guo, Fenye Bao, Jin-Hee Cho Communications

SQTrust: A New Trust Management Protocol

โ€ข What is it for?

โ€ข How to infer it?By collecting all the observations from other nodes

๐‘‡ ๐‘—๐‘ ๐‘ข๐‘ (๐‘ก)=

โˆ‘๐‘Ž๐‘™๐‘™ ๐‘–โ‰  ๐‘—

โ‘

๐‘‡ ๐‘– , ๐‘—(๐‘ก)

๐‘โˆ’1

Subjective trust

For inferring the trust belief of each node in the network

Trust Observations of node j by node i

Page 9: Integrated Social and Quality of Service Trust Management of Mobile Groups in Ad Hoc Networks Ing-Ray Chen, Jia Guo, Fenye Bao, Jin-Hee Cho Communications

SQTrust: A New Trust Management Protocol

โ€ข How to get ?

Consider the following trust metrics (namely trust components):

1. Intimacy

2. Healthiness

3. Energy

4. Cooperativeness

๐‘‡ ๐‘– , ๐‘—โ‘ (๐‘ก)=โˆ‘

๐‘‹

๐‘ค๐‘‹ร—๐‘‡ ๐‘– , ๐‘—๐‘‹ (๐‘ก)

Social Metrics

Qos Metrics

Weight of each trust component Each trust component

How to determine them?

Page 10: Integrated Social and Quality of Service Trust Management of Mobile Groups in Ad Hoc Networks Ing-Ray Chen, Jia Guo, Fenye Bao, Jin-Hee Cho Communications

SQTrust: A New Trust Management Protocol

โ€ข How to infer each trust component ?

Consider both direct trust and indirect trust.

๐‘‡ ๐‘– , ๐‘—๐‘‹ (๐‘ก )=๐›ฝ1๐‘‡ ๐‘– , ๐‘—

๐‘‘๐‘–๐‘Ÿ๐‘’๐‘๐‘ก , ๐‘‹ (๐‘ก)+๐›ฝ2๐‘‡ ๐‘– , ๐‘—๐‘–๐‘›๐‘‘๐‘–๐‘Ÿ๐‘’๐‘๐‘ก ,๐‘‹ (๐‘ก)

How to determine them?

Directly collected by Node i toward node j.

Indirect evidences given to node i by a subset of 1-hop neighbors selected.

Trade-off:

Page 11: Integrated Social and Quality of Service Trust Management of Mobile Groups in Ad Hoc Networks Ing-Ray Chen, Jia Guo, Fenye Bao, Jin-Hee Cho Communications

SQTrust: Direct Trust

โ€ข How to infer the Direct Trust of a node? Well, it depends.

- If Node i is 1-hop neighbor of node j

๐‘‡ ๐‘– , ๐‘—๐‘‘๐‘–๐‘Ÿ๐‘’๐‘๐‘ก ,๐‘‹ (๐‘ก )=๐‘‡ ๐‘– , ๐‘—

1โˆ’h๐‘œ๐‘ , ๐‘‹ (๐‘ก )

-Otherwise,

๐‘‡ ๐‘– , ๐‘—๐‘‘๐‘–๐‘Ÿ๐‘’๐‘๐‘ก ,๐‘‹ (๐‘ก )=๐‘’โˆ’๐œ†๐‘‘ โˆ† ๐‘กร—๐‘‡ ๐‘– , ๐‘—

๐‘‘๐‘–๐‘Ÿ๐‘’๐‘๐‘ก ,๐‘‹ (๐‘กโˆ’โˆ† ๐‘ก )

exponential trust decay over time.

Page 12: Integrated Social and Quality of Service Trust Management of Mobile Groups in Ad Hoc Networks Ing-Ray Chen, Jia Guo, Fenye Bao, Jin-Hee Cho Communications

SQTrust: indirect trust Inferring Indirect Trust is a little more complex.

1. Selection of Subset of 1-hop neighbors.

Threshold-based filtering: only consider trustworthy recommenders

Relevance-based trust: only consider trustworthy nodes under current trust component

<threshold

<threshold

<threshold

compromised

Low trust in healthiness

Page 13: Integrated Social and Quality of Service Trust Management of Mobile Groups in Ad Hoc Networks Ing-Ray Chen, Jia Guo, Fenye Bao, Jin-Hee Cho Communications

Trust decay over space

Trust decay over time

SQTrust: Indirect Trust 2.Calculation of indirect trust

-If there is at least one qualified neighbor:

๐‘‡ ๐‘– , ๐‘—๐‘–๐‘›๐‘‘๐‘–๐‘Ÿ๐‘’๐‘๐‘ก , ๐‘‹ (๐‘ก )=

โˆ‘๐‘šโˆˆ๐‘‰

(๐‘‡ ๐‘– ,๐‘š๐‘‹ (๐‘ก )ร—๐‘‡๐‘š , ๐‘—

๐‘‹ (๐‘ก ) )๐‘›๐‘Ÿ

-Otherwise,

๐‘‡ ๐‘– , ๐‘—๐‘–๐‘›๐‘‘๐‘–๐‘Ÿ๐‘’๐‘๐‘ก , ๐‘‹ (๐‘ก )=๐‘’โˆ’๐œ†๐‘‘ โˆ† ๐‘กร—๐‘‡ ๐‘– , ๐‘—

๐‘–๐‘›๐‘‘๐‘–๐‘Ÿ๐‘’๐‘๐‘ก ,๐‘‹ (๐‘กโˆ’โˆ† ๐‘ก )

Node iโ€™s trust in node m Node mโ€™s trust in node j

Page 14: Integrated Social and Quality of Service Trust Management of Mobile Groups in Ad Hoc Networks Ing-Ray Chen, Jia Guo, Fenye Bao, Jin-Hee Cho Communications

Model-based Evaluation

Page 15: Integrated Social and Quality of Service Trust Management of Mobile Groups in Ad Hoc Networks Ing-Ray Chen, Jia Guo, Fenye Bao, Jin-Hee Cho Communications

Model-based Evaluation

โ€ข Schema:โ€ข 1. Leverage SPN to build a semi-Markov chain

to generate the nodesโ€™ status.โ€ข 2. Reward Assignment for each status.โ€ข 3. Objective trust calculation.

Purpose: To get the objective trust as an exact global knowledge to evaluate subjective trust :

v.s.

Page 16: Integrated Social and Quality of Service Trust Management of Mobile Groups in Ad Hoc Networks Ing-Ray Chen, Jia Guo, Fenye Bao, Jin-Hee Cho Communications

a semi-Markov chain for node status

โ€ข Node Status is of 5 status representations:

โ€ข 1. Location.(int)โ€ข 2. Member.(boolean)โ€ข 3. Energy.(boolean)โ€ข 4. Healthiness .(boolean)โ€ข 5. Cooperativeness.(boolean)

trust components

To tell the position proximity of nodes

Page 17: Integrated Social and Quality of Service Trust Management of Mobile Groups in Ad Hoc Networks Ing-Ray Chen, Jia Guo, Fenye Bao, Jin-Hee Cho Communications

Location

Is fired when node moves to another region.

What is it for?1. Enable the underlying semi-Markov model to give the probability that each node is in a certain region.

2. Thus to tell whether a node is 1-hop neighbor of another.

# of tokens depends on the region a node moving into

Transition rate:

Initial speed

Wireless radio range

Page 18: Integrated Social and Quality of Service Trust Management of Mobile Groups in Ad Hoc Networks Ing-Ray Chen, Jia Guo, Fenye Bao, Jin-Hee Cho Communications

IntimacyConsider both direct trust and indirect trust.

For direct trust:1. utilize location probability of a node to infer if nodes i and j are 1-hop neighbors.

2. If they are, utilize the equation:

Based on the probability node i and node j are in the same region.

Page 19: Integrated Social and Quality of Service Trust Management of Mobile Groups in Ad Hoc Networks Ing-Ray Chen, Jia Guo, Fenye Bao, Jin-Hee Cho Communications

EnergyTo get the probability of current energy level of a node.

initialize different value to different

nodes to emphasize the heterogeneity.

-lower when node becomes uncooperative to save energy

-higher when being compromised

Initial # of tokens: depends

on the initial value

Transition rate:

Page 20: Integrated Social and Quality of Service Trust Management of Mobile Groups in Ad Hoc Networks Ing-Ray Chen, Jia Guo, Fenye Bao, Jin-Hee Cho Communications

Healthiness (CN)

A node is compromised when T_COMPRO fires

Transition rate: _๐œ† ๐‘๐‘œ๐‘š

A token goes to CN when a node is compromised

Then, either of below can happen:

1. Good-mouth a bad node with a high trust recommendation

2. Bad-mouth a good node with a low trust recommendation

Page 21: Integrated Social and Quality of Service Trust Management of Mobile Groups in Ad Hoc Networks Ing-Ray Chen, Jia Guo, Fenye Bao, Jin-Hee Cho Communications

Cooperativeness (UNCOOP)

๐‘Ÿ๐‘Ž๐‘ก๐‘’ (๐‘‡๐‘ˆ๐‘๐ถ๐‘‚๐‘‚๐‘ƒ )=๐‘“ ๐‘’ (๐ธ๐‘Ÿ๐‘’๐‘š๐‘Ž๐‘–๐‘›) ๐‘“ ๐‘’ (๐‘€๐‘‘๐‘–๐‘“๐‘“๐‘–๐‘๐‘ข๐‘™๐‘ก๐‘ฆ ) ๐‘“ ๐‘’ (๐‘†๐‘‘๐‘’๐‘”๐‘Ÿ๐‘’๐‘’ )

๐‘‡ ๐‘”๐‘

A token goes to UNCOOP when a node is uncooperative.

depends on energy, mission difficulty and neighborhood uncooperativeness degree:

Lower energy, less cooperative

Harder the mission, more cooperative

Less cooperative 1-hop neighbors, more cooperative

Group communication interval

๐‘“ (๐‘ฅ )=๐›ผ ๐‘ฅโˆ’๐œ€

Page 22: Integrated Social and Quality of Service Trust Management of Mobile Groups in Ad Hoc Networks Ing-Ray Chen, Jia Guo, Fenye Bao, Jin-Hee Cho Communications

Reward Assignment for each status

Page 23: Integrated Social and Quality of Service Trust Management of Mobile Groups in Ad Hoc Networks Ing-Ray Chen, Jia Guo, Fenye Bao, Jin-Hee Cho Communications

Objective Trust CalculationObjective trust :

๐‘‡ ๐‘—๐‘œ๐‘๐‘— (๐‘ก )=โˆ‘

๐‘‹

๐‘ค๐‘‹ โˆ™๐‘‡ ๐‘—๐‘œ๐‘๐‘— , ๐‘‹ (๐‘ก )

(1) For healthiness, energy or cooperativeness:

๐‘‡ ๐‘—๐‘œ๐‘๐‘— , ๐‘‹ (๐‘ก )=

โˆซ๐‘ก โˆ’๐‘‘โˆ†๐‘ก

๐‘ก

โˆ‘๐‘ โˆˆ ๐‘†

(๐‘Ÿ ๐‘  โˆ™๐‘ƒ ๐‘  (๐‘กโ€ฒ ))๐‘‘๐‘ก โ€ฒ

๐‘‘โˆ† ๐‘ก

(2) For intimacy:

by aggregating all the trust components calculated as:

Probability the system is at status s at time t

Page 24: Integrated Social and Quality of Service Trust Management of Mobile Groups in Ad Hoc Networks Ing-Ray Chen, Jia Guo, Fenye Bao, Jin-Hee Cho Communications

Evaluation Results

Page 25: Integrated Social and Quality of Service Trust Management of Mobile Groups in Ad Hoc Networks Ing-Ray Chen, Jia Guo, Fenye Bao, Jin-Hee Cho Communications

Evaluation Results

Parameter SettingsTotal 150 nodes, initially all are not compromised in MANETs.

Initially all are trustworthy

Based on ns3 simulation

Page 26: Integrated Social and Quality of Service Trust Management of Mobile Groups in Ad Hoc Networks Ing-Ray Chen, Jia Guo, Fenye Bao, Jin-Hee Cho Communications

Evaluation Results

Overall trust values from subjective trust v.s. objective trust

The value around 85% is the best trade-off

Page 27: Integrated Social and Quality of Service Trust Management of Mobile Groups in Ad Hoc Networks Ing-Ray Chen, Jia Guo, Fenye Bao, Jin-Hee Cho Communications

Conclusions and Future Works

1. Purpose of this paper: A protocol which minimizes the trust bias and maximize application performance.

2. Applicability: Based on the optimal protocol settings we get, we apply it for dynamic trust management with considering the environment changes.

Future Works:

Consider more sophisticated attacker behaviors, i.e. opportunistic, random and insidious attacks.

Page 28: Integrated Social and Quality of Service Trust Management of Mobile Groups in Ad Hoc Networks Ing-Ray Chen, Jia Guo, Fenye Bao, Jin-Hee Cho Communications

Thanks