INTRUSION DETECTION METHOD BASED ON GENETIC ALGORITHM AND RANDOM FOREST

In intrusion detection system, the number of original features is large, and there are redundant features and related features, which reduces the detection accuracy and increase the detection time. A genetic algorithm based on multi-layer perceptron was proposed. A four-layer perceptron neural network was established and the classification ability of the network was used as the fitness evaluation method of genetic algorithm to select the optimal feature subset. The random forest classifier was established, and the value of the hyperparameters were determined by grid verification method. The random forest classifier used the optimal feature subset to identify intrusion types. The experimental results show that the average detection accuracy of normal and 22 types of intrusion data is more than 92% on KDD99 dataset, and with a good real-time performance.