Trust Model and Anti-Attack Analysis for Social Computing

Social computing is an emerging computing paradigm that constitutes services performed by a group of people via the use of recommender systems, trust/reputation systems, social networks, peer-to-peer networks, Wikis, and online auctions. Our paper discusses social computing in trust/reputation systems and peer-to-peer (P2P) networks. Peers in P2P networks often interact with unknown or unfamiliar peers and thus some form of a trust/reputation mechanism that incorporates, say, the experience of other peers and self is needed. We evaluate the trust merits on both the feedback-only methods (direct) and the ratings (indirect) and suggest the use of aging weight, selective historic data, credibility and context to cope with malicious attacks such as colluding, white-washing and oscillation. The sparsity and cold-start problems are also discussed.