Towards robust reputation system based on clustering approach

Service computing is playing a more and more important role in current Internet activities, especially with the rapid adoption of electric markets, more and more individuals are engaging with commercial services. As the potential profit of service computing is becoming clear, malicious users are ramping up unfair rating attacks that can mislead honest service consumers into transacting with dishonest service providers. Moreover, some dishonest service providers may collude with dishonest service consumers to damage the reputation of service rivals. In this paper, we proposed a clustering-based reputation system that is robust to various unfair rating attacks. The model categorizes consumers as either honest or dishonest according to their rating ratio. It utilizes the Dirichlet distribution in determining reputation values. We analyze the profits and costs attained by the attacker and elucidate the conditions under which an attack is profitable. Experiments demonstrate that our clustering-based reputation model is more robust than the state-of-art model against currently successful attacks.