ABSTRACT
An Industrial Control System (ICS) is a computerized control system that manages and automates industrial processes. Because ICS plays a crucial role in the modern industrial infrastructure and cyber-attacks in the ICS area not only leak private information but also endanger human life and safety, how to let it be free from cyber-attack is an important issue. The cyber-attack detection that is used to detect and analyze malicious attacks is one of ICS's cybersecurity solutions. Although the existing studies and literature have proposed the corresponding solutions for ICS cyber-attacks, these methods could not capture the zero-day attacks efficiently because they all use supervised learning models. With this in mind, this study uses the one-class support vector machine (SVM) model to build an ICS cyber-attack detector using the semi-supervised learning strategy. This study also used WUSTL-IIOT 2018 dataset to train the proposed detector. The experiment result shows that (1) the proposed ICS cyber-attack detector is feasible, and (2) its performance also approaches the other detectors using supervised learning models.
