ServiceTrust: Trust Management in Service Provision Networks

Abstract

Service provision networks are popular platforms for decentralized service management. eBay and Amazon are two representative examples of enabling and hosting service provision networks for their customers. Trust management is a critical component for scaling service provision networks to larger set of participants. This paper presents Service Trust, a quality sensitive and attack resilient trust management facility for service provision networks. Service Trust has three unique features. First, it encapsulates quality-sensitive feedbacks by multi-scale rating scheme and incorporates the variances of user's behaviors into the local trust algorithm. Second, Service Trust measures the similarity of two users' feedback behavior and aggregate the local trust values into the global trust algorithm by exploiting pairwise feedback similarity scores to weight the contributions of local trust values towards the global trust of a participant. Finally, pairwise feedback similarity weighted trust propagation is utilized to further strengthen the robustness of global trust computation against malicious or sparse feedbacks. Experimental evaluation with independent and colluding attack models show that Service Trust is highly resilient to various attacks and highly effective compared to Eigen Trust, one of the most popular and representative trust models to date.