Robust Intrusion Detection System with Explainable Artificial Intelligence
Betül Güvenç Paltun, Ramin Fuladi, Rim El Malki · 2025
Machine Learning (ML) models are widely adopted for threat detection and mitigation, but their susceptibility to adversarial inputs presents new vulnerabilities, particularly in time-sensitive environments like 6G and Open Radio Access Network (O-RAN). Existing defenses, such as adversarial training, are resource intensive and often fail in real-time scenarios. To address this, we propose a novel adversarial detection and mitigation framework that leverages eXplainable Artificial Intelligence (XAI) to provide real-time insights and automated zero-touch responses. Our method is integrated into Intrusion Detection Systems (IDS) and validated through extensive testing in the Radio Resource Control (RRC) layer of the O-RAN framework. Experimental results demonstrate improved detection accuracy and reduced response time compared to baseline approaches, showcasing the effectiveness of XAI-enhanced zero-touch security in dynamic network environments.