Abstract:

In the evolving landscape of cybersecurity, large-scale dynamic networks face increasingly sophisticated threats, necessitating advanced and adaptive protection mechanisms. The integration of Artificial Intelligence (AI) with Zero Trust Architecture offers a promising path toward mitigating risks in such complex environments. However, traditional security frameworks often rely on perimeter-based defenses and centralized data models, which are insufficient in handling distributed, dynamic, and heterogeneous network structures. These limitations lead to delayed threat detection, poor scalability, and vulnerability to internal threats. To address these challenges, this paper proposes a novel framework titled Zero Trust Enhanced Threat Analytics using Federated Learning (ZETA-FL). The framework combines the principles of Zero Trust—“never trust, always verify”—with Federated Learning to enable decentralized, privacy-preserving, and real-time threat analytics. Key techniques include dynamic policy enforcement, continuous trust assessment, and AI-driven anomaly detection across nodes without centralizing sensitive data. The proposed ZETA-FL framework is designed to operate effectively across distributed network segments, ensuring secure data collaboration, minimizing latency, and enhancing threat detection accuracy. It dynamically adapts to changing threat landscapes while upholding strict data privacy requirements. Experimental evaluations demonstrate that ZETA-FL significantly outperforms traditional models in terms of detection speed, threat mitigation accuracy, and resilience against internal and advanced persistent threats. The findings underscore the framework’s potential to become a cornerstone for future-proof cybersecurity solutions in large-scale, adaptive networks.