Edge-Based Anomaly Detection: Enhancing Performance and Sustainability of Cyber-Attack Detection in Smart Water Distribution Systems

Recent advancements in cyber-physical systems have significantly increased the vulnerability of water distribution systems to cyber attacks, necessitating robust detection mechanisms. This study evaluates a decentralized detection model, leveraging edge computing to enhance the timeliness and efficiency of anomaly detection while minimizing energy consumption and carbon emissions. Despite a slight 0.4% decrease in overall performance compared to centralized systems, the decentralized model showed a 0.015% improvement in the mean time to detection, outperforming both the centralized benchmark and existing literature values. The reduction of energy consumption and emissions during training are estimated with approximately 75%, while in inference the estimate of emission reductions spans two orders of magnitude. Moreover, we demonstrate a significant advancement in attack localization, successfully assigning detected anomalies to three specific edge areas of the network. Our approach not only addresses the immediate challenges of cybersecurity in water distribution systems but also contributes to the broader goal of sustainable and resilient infrastructure.