Digital Twin Enhanced Anomaly Localization for Industrial Cyber-Physical Systems

Cybersecurity incidents such as network attacks can trigger anomalies in industrial cyber-physical systems (ICPS), leading to disastrous consequences. Accurate localization of these anomalies help in execution of precise security response measures, preventing further escalation and spread of the impact. This paper proposes a digital twin-enhanced anomaly localization (DTEAL) approach for ICPS. This DTEAL approach includes three interconnected modules: a digital twin model and a prediction compensation module, as well as an anomaly state localization module. The first module utilizes a digital twin model to predict the state of the ICPS processes and generate extended space states. Next, the prediction compensation module combines the extended data to compensate for deviations. The final state prediction results are then used to locate the anomaly state. Experimental studies conducted on a petrochemical fractionation unit simulation testbed demonstrate the effectiveness of the proposed approach.