Support Vector Machine-Based Dynamic Cyber-Attack Detection in AGC System

This paper presents novel dynamic cyber-attack detection in automatic generation control (AGC) using support vector machine (SVM). The basic idea of attack detection is based on the pattern recognition of the residual signal of the linear observer designed to estimate the states of the AGC system. Features are extracted from the residual and its derivative signal and are trained using SVM. The proposed idea is tested for various types of attack signals