Inspired by the deficiency of anonymous nodes, the descent of safety caused by the dynamics and the computation insufficiency of current trustworthiness mode, the paper proposes a computing mode of trust evaluation based on the dynamic Bayesian network. The mode calculates the trustworthiness from the direct trust and commendation one according to the historical interaction data. Time-effect factor and penalty factor are introduced in the paper to solve the problems of timeliness and malicious node. The efficiency and practicability of the mode are proved by the simulation experiment.