Trust-Preserving Overlays for Fast Reputation Aggregation in Peer-to-Peer Grid Systems∗

In recent years, Peer-to-Peer (P2P) systems and computational Grids are evolving into a new distributed computing model, called P2P Grid. This paper presents a new approach to solving the trust and security problems in a P2P Grid. Establishing trust in P2P Grids is essential to build lasting working relationships among the peers joining collective Grid applications. A P2P reputation system is thus needed to collect peer trust scores and aggregates them to yield a global reputation in both P2P and Grid systems. We use a new trust-preserving overlay network (TON) to model the trust relationships among the peers. Examining the eBay transaction trace data, we discover a power-law distribution in user feedbacks, which is proven applicable to any dynamic P2P system or any P2P Grid system. We develop a new P2P reputation system, PowerTrust, to leverage on power-law feedback characteristics. Our system is built with locality-preserving hash functions and a lookahead random walk strategy. Dynamic P2P system reconfiguration is enabled by using power nodes with well-established reputations. This power-node approach significantly reduces the global aggregation overhead. Through P2P simulation experiments on distributed file sharing and Grid parameter-sweeping applications (PSA) applications, we demonstrate the advantages of fast reputation convergence and accurate ranking of peer reputations. Simulated P2P Grid performance results are reported with enhanced query success rate, shortened job makespan, increased job success rate, and the alleviation of ill effects of malicious peers, after trust binding in either P2P systems or in P2P Grids. Keyword: Peer-to-Peer systems, Grid computing, overlay network, trust management, distributed hash table, reputation system, distributed file sharing, parameter sweeping applications (PSA, and network security. ∗ Manuscript submitted to IEEE-TPDS, Nov. 5, 2005. This work was supported by NSF Grant ITR0325409 at the University of Southern California. Corresponding author is Kai Hwang at Email: [email protected], Tel. 213 740 4470, and Fax: 213 740 4418.