Federated Learning-Based Explainable Anomaly Detection for Industrial Control Systems

We are now witnessing the rapid growth of advanced technologies and their application, leading to Smart Manufacturing (SM). The Internet Things (IoT) is one main used enable smart factories, which connecting all industrial assets, including machines control systems, with information systems business processes. Industrial Control Systems IoT-based factories top industries attacked by numerous threats, especially unknown novel attacks. As a result, distributed structure plenty IoT front-end sensing devices in SM, an effectively anomaly detection (AD) architecture for ICSs should: achieve high performance, train learn new data patterns fast time scale, have lightweight be deployed on resource-constrained edge devices. To date, most solutions not fulfilled these requirements. In addition, interpretability why instance predicted abnormal hardly concerned. this paper, we propose so- called FedeX address those challenges. experiments show that outperforms 14 other existing metrics liquid storage set. And Recall 1 F1-score 0.9857, it also SWAT proven terms training about 7.5 minutes hardware requirement memory consumption 14%, allowing us deploy tasks computing infrastructure real-time. Besides, considered as frameworks at forefront interpreting anomalies using XAI, enables experts make quick decisions trust model more.