Modeling witness trustworthiness for reliable reputation systems

Abstract

Reputation systems have been an important part for the success of online service provision systems, where a service can be, for example, an e-commerce transaction or a functional component implemented by Web service technologies. However, the existence of unfair testimonies seriously jeopardizes the reliability of reputation systems. In this thesis, we propose three approaches to address the problem of unfair testimonies. First, we propose a two-stage clustering filtering approach to filter unfair testimonies. Then we propose an integrated clustering filtering approach as an improvement of the two-stage clustering filter approach. Using a different mechanism, we further propose a novel witness trustworthiness model based on Dempster-Shafer Theory. Experimental results show that the proposed approaches can effectively mitigate the influence of unfair testimonies for reputation systems, especially for the ones supporting multi-nominal ratings, counter collusion attacks to a good extent, and eventually contribute to building reliable reputation systems.