Fuzzy Trust Recommendation Based on Collaborative Filtering for Mobile Ad-hoc Networks

Fuzzy Trust Recommendation Based on Fuzzy Trust Recommendation Based on Collaborative Filtering for Mobile Ad-hoc Collaborative Filtering for Mobile Ad-hoc NetworksNetworks

Junhai Luo1,2, Xue Liu1 , Yi Zhang 3 ,Danxia Ye2 ,Zhong Xu1

1McGill University 2University of Electronic Science and Technology of China 3University of California

September 2008

04/21/23 1

OutlineOutlineMotivationsRelated WorkArchitectureAlgorithm RealizationPerformance EvaluationConclusion and Future Work

04/21/23 2

MotivationsMotivationsMANETs characteristics:

◦Cooperative

◦Autonomous

◦Self-organized

04/21/23 3

SS

DD

ij

Motivations(cont.)Motivations(cont.)

◦Low power

◦Multi-hop

◦Vulnerable to various attacks

04/21/23 4

PDA

Pen computer

Laptop computerLaptop computer

PDA

Motivations(cont.)Motivations(cont.)

1) High trust value = ？ High or correct recommendation to other nodes.

2) Uncertain

04/21/23 5

Why？

Motivations(cont.)Motivations(cont.)Methods

◦ Collaborative filtering

◦ Fuzzy logic

04/21/23 6

Related WorkRelated WorkCONFIDANT [1]

◦ DSR (Dynamic Source Routing) with reputation systemNUGLETs [2]

◦ Virtual currencySORI [3]

◦ Secure and objective reputation schemeCORE [4]

◦ Collaborative observations and reputation mechanism

[1] S. Buchegger and J.-Y. L. Boudec, Performance analysis of the confidant protocol,” in MobiHoc ’02: Proceedings of the 3rd ACM international symposium on Mobile ad-hoc networking & computing. New York, NY, USA: ACM, 2002, pp. 226–236

[2] L. Buttyan and J.-P. Hubaux, “Nuglets: a Virtual Currency to Stimulate Cooperation in Self-Organized Mobile Ad Hoc Networks,” Tech. Rep., 2001

[3] Q. He, D. Wu, and P. Khosla, “Sori: a secure and objective reputation based incentive scheme for ad-hoc networks,” Wireless Communications and Networking Conference, 2004. WCNC. 2004 IEEE, vol. 2, pp. 825–830 Vol.2, 21-25 March 2004.

[4] P. Michiardi and R. Molva, “Core: a collaborative reputation mechanism to enforce node cooperation in mobile ad-hoc networks,” in Proceedings of the IFIP TC6/TC11 Sixth Joint Working Conference on Communications and Multimedia Security. Deventer, The Netherlands, The Netherlands: Kluwer, B.V., 2002, pp. 107–121.

04/21/23 7

ArchitectureArchitecture

04/21/23 8

i j2

jK

j1

…

K

Ri Rjk,m Rj

Ri,m

cos( , )i j

m

AlgorithmAlgorithm RealizationRealizationLocal trust Value

Collaborative filtering

Fuzzy trust recommendation

04/21/23 9

Algorithm Realization(cont.)Algorithm Realization(cont.)Local Trust Value

◦Neighbor monitoring[3]

04/21/23 10

,

( )

( )j

j mj

HF mR

RF m

( )jHF m Number of packets forwarded by node m

( )jRF m Number of packets Requested for Forwarding by node j

Algorithm Realization(cont.)Collaborative FilteringCollaborative Filtering

◦Similarity Functions

Cosine-Based Similarity

Correlation-Based Similarity

Adjusted Cosine Similarity

04/21/23 11

Algorithm Realization(cont.)Algorithm Realization(cont.)Fuzzy Method

◦Fuzzy Membership Function

◦Fuzzy Levels

◦Fuzzy Inference

04/21/23 12

Algorithm Realization(cont.)Algorithm Realization(cont.)Fuzzy Membership Function

◦ Trapezoid Membership Function (TMF)2 3

1 4

11 2

2 1

43 4

3 4

1 a x a

0 x=a or x=a

a <x<a( )

a <x<a

x ax

a a

x a

a a

04/21/23 13

a1a2 a3 a4

1

0

Trust Levels

Degree

Algorithm Realization(cont.)Algorithm Realization(cont.)Fuzzy Levels

04/21/23 14

Trust level Description Trapezoid Membership Function

HD High Distrust [-1, -0.8, -0.6]

D Distrust [-0.8, -0.6, -0.4,-0.2]

UD Undistrust [-0.4, -0.2, 0]

UT Untrust [0, 0.2, 0.4]

T Trust [0.2, 0.4, 0.6,0.8]

HT High Trust [0.6, 0.8, 1]

U Unknown [0,0,0,0]

Algorithm Realization(cont.)Algorithm Realization(cont.)Fuzzy Inference

◦Inference rule :

IF …THEN rule

For example:

IF temperature is very cold THEN turn off fan

IF temperature is very hot THEN speed up fan

04/21/23 15

Algorithm Realization(cont.)Algorithm Realization(cont.)

04/21/23 16

Start

Set node-nearest-neighbors

Retrieve node'sevaluation

Calculate thecorrelation coefficient

Calculate similarity based on fuzzy reference

Compute the trustrecommendation

End

K

Performance EvaluationPerformance EvaluationEvaluation Metrics

◦Mean Absolute Error (MAE):

Tri value of trust recommendation

Rri value of real evaluation

◦ Average Packet Drop Ratio (APDR):

04/21/23 17

1

N

i iiTr Rr

MAEN

1

1

N

DropediN

Originatedi

PacketsAPDR

Packets

Performance Evaluation(cont.)Performance Evaluation(cont.)Evaluation Setup

04/21/23 18

Parameter Value

MAC 802.11/b

Area

Speed [5,20]

Radio range 250

Placement Uniform

Movement Random waypoint

Application CBR

Sending capacity 2Mbps

Packet size 64B

Simulation time 900s

1000 1000m m

m

Performance Evaluation(cont.)Performance Evaluation(cont.)Mean Absolute Error (MAE)

04/21/23 19

NN SM Cosine Correlation Adjusted cosine

5 1.332 1.335 1.283

10 1.313 1.322 1.302

15 1.286 1.280 1.278

20 1.302 1.300 1.279

25 1.288 1.302 1.288

30 1.294 1.295 1.293

35 1.331 1.332 1.300

40 1.279 1.299 1.279

45 1.336 1.299 1.289

50 1.291 1.333 1.290

Performance Evaluation(cont.)Performance Evaluation(cont.)

04/21/23 20

Performance Evaluation(cont.)Performance Evaluation(cont.)Average Packet Drop Ratio(APDR)

04/21/23 21

Conclusion and Future WorkConclusion and Future WorkA fuzzy trust recommendation based on collaborative

filtering for MANETs.

Combining local trust and trust recommendation information based on collaborative filtering to allow nodes to represent and reason with uncertainty and imprecise information regarding other nodes' trust.

Some attack models will be done in the paper in the future.

04/21/23 22

04/21/23 23

QuestionsQuestions

?

04/21/23 24