Cloud computing environment is full of massive resources and provides resource services to the outside environment. There are lots of resource access requests both from the dynamic internal environment of cloud and the uncontrollable external environment. Traditional access control method is being widely used in cloud computing, whose rule-based method leads to huge security risk and can't fulfill the security needs of the cloud. In this paper, a dynamic risk-based access control model (DRAC) is proposed, which emphasizes the risk measurement as an auxiliary decision indicator. Moreover, a dynamic threshold of the risk is derived from the history records, and the comprehensive final decision is affect by the policy, risk measurement and dynamic threshold. To improve performance, the sliding window calculation method based on data stream is adopted. Finally, we analyze the scalability and effectiveness of the model in cloud environment application.