An Improved Support Vector Machine Attack Detection Algorithm for Industry Controls System

With the application of Internet technologies in the industrial field, industrial control systems (ICS) have the possibility of suffering malicious network attacks. Because ICS is widely used in critical infrastructure, designing attack detection algorithms for it can effectively reduce losses. Many papers have built physical models for cyber-attacks against ICS and the corresponding detection algorithms have been proposed. However, the detection performance can be further improved. This paper establishes the simulation attack on the Tennessee Eastman (TE) process and proposes an improved support vector machine (SVM) method to detect integrity attacks. This algorithm uses contribution plots and recursive feature elimination to select features. Compared with the principal component analysis (PCA) feature extraction, the detection accuracy is improved. After adding system dynamics to the algorithm, the time to detect attacks is advanced.