Securing Industry 5.0: An explainable deep learning model for intrusion detection in cyber-physical systems

Cyber-physical systems (CPS) security is critical, particularly with the advent of Industry 5.0, which seeks to revolutionize industrial ecosystems through enhanced automation, connectivity, and human-machine collaboration. While this shift promises increased efficiency and productivity, it exposes systems to advanced cyber threats. This paper introduces Cyber-Sentinet, a Deep Learning-based Intrusion Detection System (IDS) designed explicitly for CPS in industrial IoT environments to address these challenges. Unlike traditional IDS models, Cyber-Sentinet integrates Shapley Additive Explanations (SHAP) to enhance the interpretability of its decision-making process, allowing security experts to understand better and trust the system's detections. Rigorous experimentation on the Edge-IIoT-2022 dataset, which covers various cyber-attacks (e.g., DDoS, SQL injection, MITM), validates Cyber-Sentinet effectiveness. The model achieves an accuracy of 97.46 %, precision of 97.7 %, and recall of 97.2 %, with a low loss of 0.182. These results demonstrate Cyber-Sentinet ability to offer high-performance intrusion detection and valuable insights into network security, making it a robust solution for protecting Industry 5.0 CPS against sophisticated cyber threats.