PMDA: A Malicious Peer Detection Algorithm for P2P Systems

Many reputation management schemes have been introduced to assist peers in choosing trustworthy collaborators in P2P environments where honest peers coexist with malicious ones. These schemes indeed provide some useful information about the reliability of peers, but they still suffer from various attacks including slandering, collusion, etc. Consequently, it is crucial to be able to detect malicious peers. This is our focus in this paper. We first divide malicious peers into several categories, and then introduce Principal direction divisive partitioning based Malicious peer Detection Algorithm (PMDA) to identify them. Finally, we experimentally demonstrate that PMDA can efficiently and accurately detect malicious peers and hence improve the performance of the system.