Mutual Information Based Ensemble Anomaly Detection Strategy for Cyber-Physical Systems

Cyber-Physical Systems (CPS) transformed industry 4.0 by integrating the cyber world with the physical processes for smooth industrial operations. However, this integration of interconnected industrial devices with digital technologies makes CPS more susceptible to cyber and physical anomalies. To secure CPS from all such anomalies, one possible solution is the deployment of anomaly detection systems, which detect physical and cyber irregularities in the data. To this end, this study proposes a Mutual Information-based Ensemble Anomaly Detection Strategy utilizing Gradient Boosting Machine (GBM) and Light GBM to detect cyber and physical anomalies to secure CPS. The proposed methodology is evaluated on the physical and network hardware-in-the-loop dataset obtained from a Water Distribution Testbed. The proposed solution exhibited 4.6% to 6.0% and 0.7% to 1.8% accuracy improvement and a reduction of 4.7% to 6.3% and 0.7% to 1.4% in false alarm rate on the network and physical datasets respectively.