LEVERAGING ZERO TRUST ARCHITECTURE FOR NETWORK INTRUSION DETECTION: A COMPREHENSIVE EVALUATION USING THE UNSW-NB15 DATASET

Abstract

Zero Trust Architecture (ZTA) has emerged as a critical approach to enhancing cybersecurity by assuming that both internal and external network traffic must be continuously verified. This paper explores the application of ZTA principles in network intrusion detection, specifically evaluating machine learning models on the UNSW-NB15 dataset. We compare the performance of three classifiers—Random Forest (RF), Logistic Regression (LR), and XGBoost—on detecting malicious network traffic. Our results show that XGBoost achieves the highest performance with an Area Under the Curve (AUC) score of 1.00, demonstrating its effectiveness in real-time traffic monitoring. These findings prov