Gossip-based Reputation Aggregation for Unstructured Peer-to-Peer Networks

Peer-to-peer (P2P) reputation systems are needed to evaluate the trustworthiness of participating peers and to combat selfish and malicious peer behaviors. The reputation system collects locally generated peer feedbacks and aggregates them to yield global reputation scores. Development of decentralized reputation system is in great demand for unstructured P2P networks since most P2P applications on the Internet are unstructured. In the absence of fast hashing and searching mechanisms, how to perform efficient reputation aggregation is a major challenge on unstructured P2P computing. We propose a novel reputation aggregation scheme called GossipTrust. This system computes global reputation scores of all nodes concurrently. By resorting to a gossip protocol and leveraging the power nodes, GossipTrust is adapted to peer dynamics and robust to disturbance by malicious peers. Simulation experiments demonstrate the system as scalable, accurate, robust and fault-tolerant. These results prove the claimed advantages in low aggregation overhead, storage efficiency, and scoring accuracy in unstructured P2P networks. With minor modifications, the system is also applicable to structured P2P systems with projected better performance.