Decoupling Service and Feedback Trust in a P2P Reputation System

1
Decoupling Service and Feedback Trust in a P2P
Reputation System

• Gayatri Swamynathan Ben Y. Zhao Kevin C.
  Almeroth
• Department of Computer Science
• UC Santa Barbara

2
Need for Reputation in P2P Networks

• Attacks from selfish peers
• Free-riders
• Attacks from malicious peers
• Viruses, worms, Trojan horses
• Distrust among network peers
• Use of reputation to build trust
• Example eBays Feedback Forum

3
Current State of Research

• Building reputation
• Firsthand observations only
• Reliable, but inefficient
• Average lifetime ratings of a peer
• Vulnerable to attacks
• Correlated trust
• Weighted model Credibility as service provider ?
  Credibility as feedback provider
• Vulnerable to attacks like collusion

4
Current State of Research

• Open problems in reputation systems
• Unfair ratings
• Collusion
• Identity Management Sybil-proofing, etc.

5
Paper Contributions

• A novel, decoupled approach to account for a
  peers credibility as service AND feedback
  provider
• Robust to unfair ratings

Service Reputation
Reputation profile
X
Service and Feedback Reputation
Feedback Reputation
6
Related Work

• PeerTrust
• Personalized similarity measure
• Statistically hard to find common past partners
• CONFIDANT
• A node's referral is interpreted subjectively per
  node
• Not generic
• Mobile ad hoc implementation

7
The Trust Model

• Each peer OWNS two sets of reputation ratings
• A service rating (s-rating) and feedback rating
  (f-rating).
• -1,1 rating scale

let me go ahead with B
A
fAD -1
fAC 1
File (bad)
sAB -1
B
C
D
sDB 1
sCB -1
8
The Trust Model (contd.)

• Decouple service and feedback trust
• Unfair ratings
• Weigh recent feedback more heavily
• Dynamic peer personalities

9
Reputation Storage

• Orthogonal to the problem of decoupled
  reputations
• Accuracy not affected by choice of storage
• Self-storing model
• Third-party storage

10
Performance Evaluation

• Summary of results
• Effectiveness of reputation systems
• Effectiveness of decoupled approach over
  correlated trust approach

11
Simulation Environment

• Simulations in C using tools built on the SGB
  (Stanford
• Graph Base)
• Topology GT-ITM topology generator
• Peer community
• Honest
• Dishonest
• Strategic

12
Simulation Parameters
13
Experiment 1aEffectiveness against malicious
behavior

• Measuring malicious transactions in a network
  with and without our reputation model

• 40 dishonest nodes

14
Experiment 1bEffectiveness against malicious
behavior

• Measuring malicious transactions in a network
  with and without our reputation model

• 50,000 transactions

15
Experiment 1cEffectiveness against malicious
behavior

• Malicious transactions in networks with a
  conventional trust model and our decoupled model

• 50,000 transactions

16
Experiment 2aBenefits of decoupling trust

• Malicious transactions in networks with a
  conventional trust model and our decoupled model

• 40 dishonest
• Vary percentage of
• strategic nodes

17
Experiment 2bBenefits of decoupling trust

• False positives and negatives in a network with
  a conventional trust model and our decoupled
  model

• 40 dishonest
• Vary percentage of
• strategic nodes

18
Conclusions

• Proposed solution
• Robust to unfair ratings
• Dynamic peer personalities
• Open problems
• Vulnerable to collusion
• Not sybil-proof

19
Future Work

• Further analysis Trust computation error
• Comparison with PeerTrust
• Demonstrated similar results on Maze

20
Thank you!