An Intrusion Detection System for Network Cyber Security Using Hybrid Feature Selection Algorithms

One of the most important challenges of the expansion of the Internet and virtual space is cyber-attacks. These attacks are becoming new every day and it is becoming more difficult to deal with them. As a result, methods should be used to detect them, which can detect all types of cyber-attacks in the shortest possible time and with proper accuracy. Nowadays, machine learning methods are usually used to detect cyber-attacks. But since the data related to cyber-attacks have many characteristics and are kind of bulky data, as a result, the accuracy of conventional machine learning methods to detect them is usually low. In this research, we have used a hybrid feature selection method to select optimal features from the database related to cyber-attacks, which increases the accuracy of attack detection by classification models. In the proposed feature selection method, first the features that have the least redundancy with each other and at the same time are most related to the category variables (labels) are selected by the MRMR algorithm. Then, using a wrapper feature selection method based on the gray wolf optimization (GWO) algorithm to select a subset of the features selected from the previous step, which maximizes the accuracy of the SVM classifier model, is used this subset has optimal features by which the SVM model is trained. As a result, the accuracy of detecting cyber-attacks by the SVM model increases. According to the simulation results, the average accuracy of the proposed method for detecting cyber-attacks is 99.84%, which has improved compared to the intrusion detection methods of the reference article.