Abstract:

Reputation management is a crucial task in Peer-to-Peer networks, social networks and other decentralized dis- tributed systems. In this work we investigate the role of users’ negative opinions in order to devise fully de- centralized reputation management mechanisms. Our study is motivated by the limitations of methods pro- posed in literature that are based on the idea of propa- gating positive opinions, most notably EigenTrust [9], a cornerstone method for reputation management in de- centralized systems. EigenTrust makes use of a transi- tive definition of trust: a peer tends to trust those peers who have a high reputation in the opinion of trustworthy peers. While EigenTrust has been shown to be effective against a number of threat attacks from coalitions of malicious players, it does not address properly more so- phisticated threat attacks. In this paper we propose a new approach to the design of fully decentralized reputation mechanisms that com- bine negative and positive opinions expressed by peers to reach a global consensus on trust and distrust values for each peer of the network. We show how these strategies outperform EigenTrust in terms of number of successful transactions against a large set of sophisticated threat attacks posed by coalitions of malicious peers. We also discuss a clustering method that achieves detecting most of the malicious peers with high precision.