TRIBE: Trust Revision for Information Based on Evidence

In recent years, the number of information sources available to support decision-making has increased dramatically. However, more information sources do not always mean higher precision in the fused information. This is partially due to the fact that some of these sources may be erroneous or malicious. Therefore, it is critical to asses the trust in information before performing fusion. To estimate trust in information, existing approaches use trustworthiness of its source as a proxy. We argue that conflicts between information may also serve as evidence to reduce trust in information. In this paper, we use subjective opinions to represent information from diverse sources. We propose to exploit conflicts between opinions to revise their trustworthiness. For this purpose, we formalise trust revision as a constraint optimisation problem. Through extensive empirical studies, we show that our approach significantly outperform existing ones in the face of malicious information sources.