A context-aware trust establishment and mapping framework for web applications

In order to reflect the dynamic feature of situations and of peers' behavior more accurately, context information, which contains different attributes of situations, should be considered explicitly when evaluating trust. A context-aware trust establishment framework based on Bayesian networks is proposed. In our proposal, the contextual factors are inferred statistically, not assigned in ad hoc way as most existing solutions. We also propose a Bayesian trust mapping approach between related contexts to help application agents infer the trustworthiness in an unfamiliar context from the trust information in other contexts. We argue that the relevance of contexts is specific to agents and it should not be estimated independent of agents' trust or even fixed as in other frameworks. Our approach makes the relevance measurement implicit together with trust mapping through Bayesian networks. Simulation experiments show that our system can make more accurate trust inferences than ad hoc context-aware systems.