GroupRep: A Robust Group-based Reputation System for Peer-to-Peer Networks

Reputation systems, which allow peers to collaborate on estimating the trustworthiness of other peers, are considered as one of the promising directions against malicious behaviors in large scale Peer-to-Peer networks. However, a weakly designed reputation system is vulnerable to attacks such as unfair rating and collusion. This paper presents GroupRep a novel group-based reputation system for Peer-to-Peer networks. In the GroupRep model, peers are allowed to form social groups, in which member peers will collaborate against attacks on the reputation system. For management of the group, a set of robust and light-weighted protocols are proposed based on distributed voting. Experiments show that the GroupRep model is effective and inexpensive, and the system is robust enough even under the situation that the majority of the peers are malicious and they cooperate deliberately to subvert the system.