An Intrusion Detection Mothod based on Feature Selection and Binary Classification Grouped Learning
Xueqin Gao, Tao Wang, Qian Wu, Jiehu Wu · 2022 IEEE 6th Advanced Information Technology, Electronic and Automation Control Conference (IAEAC ) · 2022
The detection rate of the minority class is generally low because of the unbalanced distribution in the intrusion dataset. To improve the detection rate of the minority classes in the intrusion detection dataset, this paper proposes an intrusion detection method (GLM-IDS) based on the binary classification and ensemble learning method. It first selects the features that are important for the binary classification training based on the features importance method. To overcome the difficulty during the multi classification training which cannot give attention to the minority classes, this method transforms the multi classification to the binary classification problem and then performs binary classification training in groups to keep the balance of the minority classes in the training dataset. The final multi classification decision is computed from the multi binary classification results based on the ensemble learning method. The experiment of GLM-IDS is compared to other multi classification methods such as machine learning and deep learning models. The result shows the feasibility that this method improves the detection rate of the U2R and R2L which belong to the minority classes without reducing the detection performance of the majority classes.