Gossip-Based Reputation Management for Unstructured Peer-to-Peer Networks*

To build an efficient reputation system for peer-to-peer (P2P) networks, we need fast mechanisms to aggregate peer evaluations and to disseminate updated scores to a large number of peer nodes. Unfortunately, unstructured P2P networks are short of secure hashing and fast lookup mechanisms as in structured P2P systems like the DHT-based Chord. In light of this shortcoming, we propose a gossiping mechanism to support unstructured P2P communications. This paper reports the design principles and performance results of a GossipTrust system. Gossip-based communications are shown effective to support fast peer scoring and distribution. Gossiping enables light-weight aggregation of local scores and spreading of updated scores to all nodes in O(log2 n) time, where n is the P2P network size. The GossipTrust works as follows: Each peer gossips with a few randomly selected other nodes to share reputation data, periodically. Gossip-based protocols do not require error recovery and thus enjoy the simplicity with a moderate processing overhead. Bloom filters are specially tailored for storing and retrieving global reputation rankings on each node. We apply identity-based cryptography to secure gossiped reputation data. We report simulation results on distributed file-sharing and parameter-sweeping applications on GossipTrust. 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 even better performance, which is beyond the scope of this paper.