The traditional authentication technology showed some deficiencies in cloud computing, in order to decrease intrusion of untrustworthy user, this paper discusses a scheme for user behavior authentication in cloud computing environment which combines traditional authentication and behavior authentication. The paper presents the creation of the user behavior authentication set. The paper presents the realization process of the whole mechanism. The paper proposes Bayesian network model to predict the level of user behavior authentication based on the combination of the evidence of past transaction and real-time user behavior information. We give an example to demonstrate the effectiveness of the prediction model. This research is of theoretically and practically significant for enhancing user authentication and decreasing intrusion of untrustworthy user effectively.