SCADA Data‐Driven Anomaly Detection
Abdulmohsen Almalawi, Zahir Tari, Adil Fahad, Xun Yi · 2020
This chapter introduces a novel unsupervised Supervisory Control and Data Acquisition (SCADA) Data-driven Anomaly Detection approach (SDAD), generating from unlabeled SCADA data, proximity anomaly-detection rules based on the clustering method. It describes consistent/inconsistent observations for SCADA points and the two methods that contribute to the development of SDAD: a method that separates inconsistent observations from consistent ones of multivariate SCADA points and a method that extracts proximity-based detection rules, which is used to detect inconsistent observations. The chapter focuses on the setup of the experimental environment in order to evaluate the robustness of the proposed approach. It aims to evaluate the accuracy of SDAD. Two parts in this approach are evaluated: the first part is the accuracy of the separation between consistent and inconsistent observations as a first phase to extract detection rules, while the second part is the detection accuracy of these extracted detection rules.