Reputation Systems via Belief Propagation

Trust and reputation mechanisms have various application areas from online services to mobile ad-hoc networks (MANETs). Most well-known commercial websites such as eBay, Amazon, Netflix and Google use some types of reputation mechanisms. Despite recent advances in reputation systems, there is yet a need to develop reliable, scalable and dependable schemes that would also be resilient to various ways a reputation system can be attacked. We approach the reputation management problem as an inference problem and describe it as computing marginal likelihood distributions from complicated global functions of many variables. However, we observe that computing the marginal probability functions is computationally prohibitive for large scale reputation systems. Therefore, we propose to utilize the belief propagation algorithm to efficiently compute these marginal probability distributions, resulting in a belief propagation-based iterative trust and reputation management approach (BP-ITRM). Compared with some well-known reputation management techniques, e.g., Averaging Scheme, Bayesian Approach, and Cluster Filtering, BP-ITRM is superior in terms of robustness against common attacks such as ballot-stuffing and bad-mouthing. Finally, BP-ITRM introduces only a linear complexity in the number of service providers and consumers, and thus is suitable for deployment in large-scale systems.