ICS intrusion detection system and method fusing reinforcement learning and feature selection optimization

The invention discloses an ICS intrusion detection system and method fusing reinforcement learning and feature selection optimization, and the method comprises the steps: carrying out the binary coding and population initialization of the data feature selection of a historical data set of an industrial control system (ICS), carrying out the offline training through SVM-reinforcement learning, taking the accuracy obtained on a verification set as a fitness function, and carrying out the feature selection optimization through SVM-reinforcement learning. Designing crossover operation and mutation operation based on cumulative probability to update the population, and obtaining an optimal feature set after iterative optimization; and performing feature selection on the ICS real-time data set based on the optimal feature set, and performing online intrusion detection test on the real-time data set through support vector machine (SVM)-reinforcement learning so as to obtain an intrusion detection performance index.