I dentification and A ccess C ontrol of A rtificial I ntelligence in Z ero T rust N etwork A rchitecture

With the rapid development of cloud computing, mobile office, and Internet of things, the boundary of enterprise network is increasingly blurred, and the traditional security model has been unable to deal with the threat of today's complex network. With the core principle of "never trust, always verify", zero trust architecture (ZTA) breaks the traditional security model based on network boundaries. Artificial intelligence technology provides key support for the intelligent upgrading of zero trust architecture. This paper systematically reviews the research progress in this field from four dimensions: continuous authentication, multi-source anomaly detection, dynamic trust evaluation and adaptive access control. According to the research, the AUC of the multimodal CNN-LSTM hybrid model can reach 0.92, and the federated learning takes into account both privacy protection and detection accuracy; The behavior feature fusion framework based on XGboost can effectively identify internal threats. Device activity and email attachment behavior are the key distinguishing features; The subjective logic method is highly interpretable, and the quantum inspired model provides a new tool for modeling the uncertainty of trust; At present, it still faces three major challenges: privacy protection, insufficient model interpretability and standardized interoperability. This paper establishes a four-dimensional analysis framework of AI enabling zero trust, which provides a reference for subsequent theoretical research and engineering deployment.