Research on Trust Propagation Models in Reputation Management Systems

Feedback based reputation systems continue to gain popularity in eCommerce and social media systems today and reputation management in large social networks needs to manage cold start and sparseness in terms of feedback. Trust propagation has been widely recognized as an effective mechanism to handle these problems. In this paper we study the characterization of trust propagation models in the context of attack resilience. We characterize trust propagation models along three dimensions: (i) uniform propagation and conditional propagation, (ii) jump strategies for breaking unwanted cliques, and (iii) decay factors for differentiating recent trust history from remote past history. We formally and experimentally show that feedback similarity is a critical measure for countering colluding attacks in reputation systems. Without feedback similarity guided control, trust propagations are vulnerable to different types of colluding attacks.