As society evolves, people have become increasingly reliant on the internet, and network access control technologies ensure the security of network environments. However, traditional network access control technologies generally suffer from limitations such as static verification cycles, single authorization methods, and a lack of continuous behavior tracking, making it difficult to address dynamic and complex network threats. To address this, the study proposes a network access control model based on the zero-trust security concept, integrating dynamic authorization verification methods to keep user verification in a continuous and dynamic state, thereby enhancing the system’s overall protective capabilities. Through simulation experiments, the results show that compared to KNN, SVM, and MLP models, the proposed ZT-DA model achieves an accuracy rate of 98.4% and a false positive rate of 3.8% when accessing 400 times, with clearer classification boundaries, enabling more accurate distinction between normal and abnormal behavior. When the sample size is 200, the trust calculation times for different stages are 78 ms, 65 ms, and 70 ms, respectively. In practical applications, the model achieved an interception rate of 100%, a false interception rate of 0%, and a control range of 100%. The research results indicate that the model can monitor threat behaviors in real time, effectively intercept malicious attacks, and enhance information security protection capabilities with a lower false positive rate while ensuring a smooth user network experience.