Search

This study introduces a novel, practical approach for designing a hierarchical online anomaly detection system for industrial cyber-physical systems. The proposed method utilizes the Hierarchical Temporal Memory (HTM) unsupervised learning algorithm, which requires data to be encoded as sparse binary distributed representations (SDRs). A new SDR encoding method termed the temporal sequence encoder (TSSE) is presented to convert system outputs into SDRs. This approach enables HTM to retain high memory capacity and robust performance when processing data streams of slowly varying physical measurements, typical of many industrial processes. The effectiveness of the proposed system is demonstrated on the Secure Water Treatment (SWaT) dataset, which comprises data collected from a fully operational, scaled-down water treatment plant. The system achieves a recall of 0.906, a precision of 0.935, and an F1 score of 0.92 on SWaT. Compared to previous methods, our approach achieves state-of-the-art recall (~5.3% improvement), along with competitive precision and F1 score, by learning in an online manner without the need for expensive dataset collection, labeling, or retraining phases. These findings suggest that the proposed online anomaly detection method can be effectively applied to a broad range of water treatment and large-scale industrial cyber-physical systems. [ABSTRACT FROM AUTHOR]