Anomaly Detection in Predicted Water Treatment Data using Hybrid CNN-LSTM Network Model

Abstract

Water treatment systems are among the industrial control systems where it is essential to detect anomalies accurately and efficiently due to the potential threat to public health. With advances in computer science, machine learning models have been successfully used in the anomaly detection process in recent years. In this paper, a hybrid CNN-LSTM network model is proposed to detect anomalies in water systems. Using a statistical window-based anomaly detection approach, the performance of the proposed model in detecting different types of attacks is analyzed on the open SWaT dataset. Experimental results show that the precision, recall and F1-score values of the proposed model are 0.994, 0.973 and 0.983, respectively, and can be successfully used to detect anomalies in water treatment systems.