Distributed Real-Time Anomaly Detection in Networked Industrial Sensing Systems

Reliable real-time sensing plays a vital role in ensuring the reliability and safety of industrial cyber-physical systems (CPSs) such as wireless sensor and actuator networks. For many reasons, such as harsh industrial environments, fault-prone sensors, or malicious attacks, sensor readings may be abnormal or faulty. This could lead to serious system performance degradation or even catastrophic failure. Current anomaly detection approaches are either centralized and complicated or restricted due to strict assumptions, which are not suitable for practical large-scale networked industrial sensing systems (NISSs), where sensing devices are connected via digital communications, such as wireless sensor networks or smart grid systems. In this paper, we introduce a fully distributed general anomaly detection (GAD) scheme, which uses graph theory and exploits spatiotemporal correlations of physical processes to carry out real-time anomaly detection for general large-scale NISSs. We formally prove the scalability of our GAD approach and evaluate the performance of GAD for two industrial applications: building structure monitoring and smart grids. Extensive trace-driven simulations validate our theoretical analysis and demonstrate that our approach can significantly outperform state-of-the-art approaches in terms of detection accuracy and efficiency.