Sources and Monitors: A Trust Model for Peer-to-Peer Networks

In this paper, we introduce an objective model of trust in peer-to-peer networks. Based on this model, we develop protocols that can be used by the peers in a peer-to-peer network to compute the trust values of other peers in these networks. According to our model, the trust value of a peer is the probability that this peer sends correct messages to other peers, provided that this probability is at least 0.6. (A peer whose probability of sending correct messages is less than 0.6 is regarded as a bad peer that cannot be trusted by other peers in the network.) Each peer actively monitors several good peers in the network and accurately estimates the trust values of each of them. The peers then exchange messages about the trust values of the good peers that they have monitored, and each of them ends up accurately computing the trust values of many good peers in the network, even though many of the exchanged messages are arbitrarily wrong. Through analysis and simulation, we show that a peer in a network can compute the trust values of about 100 good peers in the network, while keeping the error in computing these trust values below 10^-4.