With the explosive growth of the variety and quantity of multimedia information in the Internet of Things (IoT) environment, its security problem is becoming more and more prominent. Therefore, this paper constructs APODAC dynamic access control model. The information processing of massive data of IoT is carried out through the fusion technology of multiple media features. Based on the real-time access behavior sequence of IoT, a fuzzy reasoner is used to analyze the degree of risk and assess the network security posture. Based on the degree of risk, IoT access rights are dynamically adjusted. The simulation experiment results show that the fuzzy reasoning method in this paper has a 4.4% higher risk detection rate for IoT network and a 10.5% decrease in false alarm rate compared to the traditional SVM method. In risk behavior oriented dynamic access control, the APODAC model proposed in this paper still outperforms the other 2 models in terms of response time for both higher number of access requests and smaller amount of access request data.