Abstract: A trust management framework based on Bayesian networks is proposed to reflect the dynamic and uncertain feature of entities' behaviors in open network environment more accurately. The context information is considered explicitly and its effect on entity behaviors is inferred statistically. A Bayesian trust mapping approach between the related contexts is also proposed, which allows the application agents to infer the trustworthiness in an unfamiliar context from the trust information in other contexts. Simulation results showed that the system can make more accurate trust inferences than ad hoc context-aware systems.