Trust and Reputation System

1. Trust and Reputation System S. Felix Wu University of California, Davis wu@cs.ucdavis.edu http://www.cs.ucdavis.edu/~wu/ Trust and Reputation System

2. OCC, TSO, 2PL • T1 r X • T1 r Y • T1 w X • T1 r Z • T1 w Y Trust and Reputation System

3. Trust in P2P • The Service Provider provides a management system for trust and reputation • Google’s “PageRank” • Antivirus system • eBay’s seller reputation system • PKI • P2P -- everything hopefully to be P2P • Decentralized model for trust Trust and Reputation System

4. Cheating & Incentives • Selfish users in Gnutella and Bittorrent • eBay flaw seller ranking • Google page rank • Selfishness or Reputation boost Trust and Reputation System

5. P2P Trust Model • Less vulnerable? • Harder to implement? In a decentralized setting? Trust and Reputation System

6. Problem • Problem: • Reduce inauthentic files distributed by malicious peers on a P2P network. • Motivation: “Major record labels have launched an aggressive new guerrilla assault on the underground music networks, flooding online swapping services with bogus copies of popular songs.” -Silicon Valley Weekly Trust and Reputation System

7. Problem 0.9 0.1 • Goal: To identify sources of inauthentic files and bias peers against downloading from them. • Method: Give each peer a trust value based on its previous behavior. Trust and Reputation System

8. Some approaches • Past History • Friends of Friends • EigenTrust • PeerTrust • TrustDavis Trust and Reputation System

9. Terminology t3=.5 C12=0.3 C23=0.7 t1=.3 C21=0.6 t2=.2 C14=0.01 t4=0 Peer 3 • Local trust value:cij.The opinion that peer i has of peer j, based on past experience. • Global trust value: ti.The trust that the entire system places in peer i. Peer 1 Peer 2 Trust and Reputation System Peer 4

10. Local Trust Values • Each time peer i downloads an authentic file from peer j, cij increases. • Each time peer i downloads an inauthentic file from peer j, cij decreases. Cij= Peer i Peer j Trust and Reputation System

11. Normalizing Local Trust Values Peer 1 C12=0.9 Peer 2 C14=0.1 Peer 4 Peer 4 Peer 2 Peer 1 • All cij non-negative • ci1 + ci2 + . . . + cin = 1 Trust and Reputation System

12. Local Trust Vector Peer 1 C12=0.9 C14=0.1 Peer 2 Peer 4 Peer 4 c1 Peer 2 Peer 1 • Local trust vector ci:contains all local trust values cij that peer i has of other peers j. Trust and Reputation System

13. Past history ? ? ? ? ? Peer 6 ? Peer 4 Peer 1 • Each peer biases its choice of downloads using its own opinion vector ci. • If it has had good past experience with peer j, it will be more likely to download from that peer. • Problem: Each peer has limited past experience. Knows few other peers. Trust and Reputation System

14. Friends of Friends Peer 6 Peer 4 Peer 1 Peer 2 Peer 8 • Ask for the opinions of the people who you trust. Trust and Reputation System

15. Friends of Friends Peer 4 Peer 1 Peer 2 Peer 8 Peer 4 • Weight their opinions by your trust in them. Trust and Reputation System

16. The Math What they think of peer k. And weight each friend’s opinion by how much you trust him. Ask your friends j .1 .5 0 0 0 .2 .1 .3 .2 .3 .1 .1 0 .2 0 .3 0 .5 .1 0 0 0 .2 Trust and Reputation System

17. Problem with Friends • Either you know a lot of friends, in which case, you have to compute and store many values. • Or, you have few friends, in which case you won’t know many peers, even after asking your friends. Trust and Reputation System

18. Dual Goal • We want each peer to: • Know all peers. • Perform minimal computation (and storage). Trust and Reputation System

19. Knowing All Peers • Ask your friends: t=CTci. • Ask their friends: t=(CT)2ci. • Keep asking until the cows come home: t=(CT)nci. Trust and Reputation System

20. Minimal Computation • Luckily, the trust vectort, if computed in this manner, converges to the same thing for every peer! • Therefore, each peer doesn’t have to store and compute its own trust vector. The whole network can cooperate to store and compute t. Trust and Reputation System

21. Non-distributed Algorithm • Initialize: • Repeat until convergence: Trust and Reputation System

22. Distributed Algorithm .1 .5 0 0 0 .2 .1 .3 .2 .3 .1 .1 0 .2 0 .3 0 .5 .1 0 0 0 .2 • No central authority to store and compute t. • Each peer i holds its own opinions ci. • For now, let’s ignore questions of lying, and let each peer store and compute its own trust value. Trust and Reputation System

23. Distributed Algorithm For each peer i { -First, ask peers who know you for their opinions of you. -Repeat until convergence { -Compute current trust value: ti(k+1) = c1jt1(k) +…+ cnjtn(k) -Send your opinion cij and trust value ti(k)to your acquaintances. -Wait for the peers who know you to send you their trust values and opinions. } } Trust and Reputation System

24. Probabilistic Interpretation Trust and Reputation System

25. Malicious Collectives Trust and Reputation System

26. Pre-trusted Peers • Battling Malicious Collectives • Inactive Peers • Incorporating heuristic notions of trust • Convergence Rate Trust and Reputation System

27. Pre-trusted Peers • Battling Malicious Collectives • Inactive Peers • Incorporating heuristic notions of trust • Convergence Rate Trust and Reputation System

28. Secure Score Management M ? ? M M M ? ? • Two basic ideas: • Instead of having a peer compute and store its own score, have another peer compute and store its score. • Have multiple score managers who vote on a peer’s score. Score Manager Distributed Hash Table Score Managers Trust and Reputation System

29. PeerTrust System Architecture Trust Manager Trust Evaluation Feedback Submission P1 Trust Data P6 P2 Data Locator P2P Network P2P Network P5 P3 P4 Trust and Reputation System

30. How to use the trust values ti • When you get responses from multiple peers: • Deterministic: Choose the one with highest trust value. • Probabilistic: Choose a peer with probability proportional to its trust value. Trust and Reputation System

31. Load Distribution Probabilistic Download Choice Deterministic Download Choice Trust and Reputation System

32. Threat Scenarios • Malicious Individuals • Always provide inauthentic files. • Malicious Collective • Always provide inauthentic files. • Know each other. Give each other good opinions, and give other peers bad opinions. Trust and Reputation System

33. More Threat Scenarios • Camouflaged Collective • Provide authentic files some of the time to trick good peers into giving them good opinions. • Malicious Spies • Some members of the collective give good files all the time, but give good opinions to malicious peers. Trust and Reputation System

34. Malicious Individuals Trust and Reputation System

35. Malicious Collective Trust and Reputation System

36. Camouflaged Collective Trust and Reputation System

37. P2P Electronic Communities Trust and Reputation System

38. Motivation Trust and Reputation System

39. Motivation • Should we buy? • How do we decide? Trust and Reputation System

40. Motivation Trust and Reputation System

41. Motivation • Should we buy? • How do we decide? • What we want: • accurately estimate risk of default • minimize the risk of default • minimize losses due to pseudonym change • avoid trusting a centralized authority • How do we achieve these goals? Trust and Reputation System

42. Motivation • TrustDavis is a reputation system that realizes these goals. • It recasts these goals as the following properties: Trust and Reputation System

43. Motivation • Agents can accurately estimate risk • Third parties provide accurate ratings • Honest buyer/seller avoids risk (if possible) • Insure transactions • No advantage in obtaining multiple identities • Agents can cope with pseudonym change • No need to trust a centralized authority • No centralized services needed Trust and Reputation System

44. Motivation Incentive Compatibility: Each player should have incentives to perform the actions that enable the system to achieve a desired global outcome. Trust and Reputation System

45. Motivation • Agents can accurately estimate risk • Third parties provide accurate ratings • Honest buyer/seller avoids risk (if possible) • Insure transactions • No advantage in obtaining multiple identities • Agents can cope with pseudonym change • No need to trust a centralized authority • No centralized services needed Incentive Compatibility! Trust and Reputation System

46. Motivation $100 A B C A Reference is: Acceptance of Limited Liability. Trust and Reputation System

47. Motivation • Agents can accurately estimate risk • Third parties provide accurate ratings • Parties are liable for the references they provide • Honest buyer/seller avoids risk (if possible) • Insure transactions • Buyers/sellers pay for references to insure their transactions • No advantage in obtaining multiple identities • Agents can cope with pseudonym change • References are issued only to trusted identities • No need to trust a centralized authority • No centralized services needed • Anyone can issue a reference Use References! Trust and Reputation System

48. Outline • TrustDavis leverages social networks • For now, examples assume No False Claims (NFC) • The use of TrustDavis does NOT preclude trade outside the system. Trust and Reputation System

49. Paying for References 50 150 100 50 150 Trust and Reputation System

50. Paying for References $100 each Trust-me.com Blowout SALE! $50 each! $150! How much is vb willing to pay to insure the transaction? (No riskless profitable arbitrage criterion) Example: • vb wants to buy three shirts. • Shirts cost $100 each from a trustworthy seller • Unknown seller offers shirts for $50 each (but maybe they are only worth $25). • vb would risk 3 x $50 = $150 in the transaction • vb can borrow and lend money at rate r=1.25 through the period of the transaction For $30, vb can insure herself! Trust and Reputation System