The open environment in cloud computing is much more complex and unpredictable

so how to identify untrustworthy cloud end-user by analyzing abnormal user behaviors is an important topic in cloud security.This paper proposes a model for behavior analysis based on incomplete information multi-stage dynamic games

in which current action and historical action

false negative and false positive in network detection methods are considered to improve the accuracy and comprehensiveness of the dynamic judgment of end-use trustworthiness.The experimental results show that it can discriminate potential untrustworthy cloud end-user

and decrease intrusion effectively while perfect Bayesian equilibrium is reached

laying the foundation for active safety mechanism.