Digital Twin-Assisted anomaly detection for industrial scenarios

Abstract

Industry 5.0 is the current industrial paradigm that inherits the technological diversity of its predecessor, Industry 4.0, but includes three priority goals: (i) resilience, (ii) sustainability and (iii) human-centeredness. Through these three goals, Industry 5.0 pursues a more far-reaching digital transformation in industrial ecosystems with high protection guarantees. However, the deployment of innovative information technologies for this new digital transformation also requires considering their implicit vulnerabilities and threats in order to avoid any negative impacts on the three Industry 5.0 goals, and to prioritize cybersecurity aspects so as to ensure acceptable protection levels. This paper, therefore, proposes a detection framework composed of a Digital Twin (DT) and machine learning algorithms for online protection, supporting the resilience that Industry 5.0 seeks. To validate the approach, this work includes several practical studies on a real industrial control testbed to demonstrate the feasibility and accuracy of the framework, taking into account a set of malicious perturbations in several critical sections of the system. The results highlight the effectiveness of the DT in complementing the anomaly detection processes, especially for advanced and stealthy threats.