A New Feature Selection Approach and Classification Technique for Current Intrusion Detection System

These days, various devices including computers, smartphones, internet of things (IoT), and cloud services are using computer networks for data communications. As the computer network is being used extensively, it becomes the target of many attacks. It can be different attacks such as denial of service attack (DoS), remote to user attack (R2L), user to remote attack (U2R), and probing attack. To protect communication networks from network-based attacks, intrusion detection systems (IDSs) have been proposed in many studies. However, today IDSs are not good enough to detect new attack types in the communication networks. To increase the efficiency of the current IDSs, a subset of features needs to be obtained before performing the machine learning classifiers. In this study, a new feature selection method is proposed for current IDSs. In addition, the proposed method is combined with machine learning classifiers and tested on KDD ’99 dataset and %99.81 accuracy rate was obtained. The obtained performance is pretty high to separate network attacks from the normal traffic.