Group trust yields improved scalability and anomalydetection for p2p systems

Abstract

This paper implements an existing computational model of trust and reputationapplied to a P2P environment, and extends the approach using a novel group trust calculationthat demonstrates improved scalability and anomaly detection for P2P systems. Our analysis isbased on results obtained by simulating a P2P environment using the JXTA open source platform.A trust and reputation model was implemented in the same platform, allowing to constructinga baseline for the behavior of the nodes using combined trust and reputation coefficients in ascenario without malicious nodes. Then simulations were conducted with malicious nodes andthe effect of trust and reputation factors were analyzed regarding their influence on the anomalydetection capacity and scalability in P2P communications. Several simulation scenarios wereconfigured and explored, considering the presence of different number of malicious nodes in theP2P environment, with both constant and variable behavior. Other scenarios included calculationsof combined trust and reputation for node groups. The results show that group trust ensure moreinteractions among nodes, even in the presence of a large number of malicious nodes (60% of thetotal), besides providing focused identification of malicious nodes inside groups.