AI‐Enabled Zero Trust Model for Cyberattack Mitigation in Decentralized 6G Network

ABSTRACT Another trend in the development of the sixth‐generation (6G) networks is new and unprecedented features, including ultra‐low latency, very high device density, autonomous communication, and distributed intelligence. Nevertheless, the resulting improvement significantly expands the cyberattack surface, rendering traditional perimeter‐based defense mechanisms ineffective. As a solution to this problem, we introduce an AI‐Based Zero Trust Architecture to operate in decentralized 6G settings. The framework gets rid of implicit trust through the introduction of active checking, adaptive access control, and the response to anomalies in real time. Graph‐Based Temporal Convolutional Network is used to detect the multi‐stage cyberattack, and it both captures the spatial traffic relationship and temporal event behavior. Multi‐Agent Deep Deterministic Policy Gradient reinforcement learning makes the optimization of policy decisions optimal and hopes to have a collaborative response to threats but independent among the distributed nodes. A Spiking Neural Network achieves fast, low‐power mitigation that can benefit resource‐constrained edge machines to ensure that long‐term cryptographic resilience is realized by having trust validation performed using Lattice‐Based Polynomial Commitment. The system has 98.5% detection, 96.8% F1 score, 0.982 AUC, and a false alarm rate of 3.1%, which is much better than the current models such as federated learning, DNN, SVM, and rule‐based IDS. These findings clearly indicate that the proposed model is potent and appropriate in its use as a scalable Zero Trust model in safeguarding future 6G networks.