Optimal Sensor Data Resampling for Anomaly Detection in Industrial Control Systems

The goal of this study is to identify optimal sensor data sampling rates for anomaly detection in industrial control systems (ICS). The identification of optimal sensor data sampling rates allows ICS operators to store less data while training adequate anomaly detection models. In this study, the autoencoder method was utilized to enhance the effectiveness of the detection process and statistical (histogram) anomaly scores were applied to determine anomaly threshold points. The experiment on the proposed method was conducted using the Secure Water Treatment (SWaT) and the HIL-based Augmented ICS Security (HIL-HAI) datasets. The results of the experiment demonstrate that the proposed method outperformed the current anomaly detection techniques. Our code is publicly available at https://github.com/Ermiyas21/ICSSensor .