The Tell-Tale Neighbors: Trust and Reputation in Peer-to-Peer Systems

Peer-to-peer systems are an attractive means of sharing data and services. However, the problem of how to efficiently decide which peers are to be trusted still remains unsolved. In this work we propose a decentralized trust management system based on reputation, for unstructured, self-organizing peer-to-peer networks. Our protocol takes advantage of the unstructured nature of the network to render malicious behavior like lying and colluding risky. The reputation information of each peer is stored in its neighbors and piggy-backed on its query-hits. By simulating the behavior of networks both using and not using a rating scheme we were able to show that just a few dishonest nodes can flood the network with false results, whereas this phenomenon is virtually eliminated when using a rating scheme.