AI-Driven Intrusion Detection and Threat Mitigation in Zero Trust Architectures

Abstract

This study explores the role of blending Artificial Intelligence (AI) with Zero Trust Architecture (ZTA) to boost Cybersecurity through a solid Intrusion Detection System (IDS). Using a design science and postpositivist approach, the research crafts and assesses AI-driven IDS models such as Decision Tree, Random Forest, Long Short-Term Memory (LSTM), and Transformer by leveraging the University of New South Wales - Network Benchmark 2015 (UNSW-NB15) dataset within a simulated ZTA environment. The study evaluates how well these models perform based on metrics like accuracy, precision, recall, and F1-score to pinpoint the best architecture for spotting both insider and external threats. The results show that the Transformer-based IDS outshines the others with an impressive 99.2% accuracy rate, proving its exceptional ability in real-time anomaly detection and threat classification. These findings highlight that merging deep learning with ZTA’s identity-based access control significantly boosts network resilience and helps reduce lateral movement threats. This research presents an innovative AI-ZTA integrated model designed for proactive, adaptive, and scalable cybersecurity defense. It also offers valuable insights for organizations aiming to strengthen digital trust and policy frameworks in the face of increasingly sophisticated cyber threats.