Influence Analysis of Feature Selection to Network Intrusion Detection System Performance Using NSL-KDD Dataset

Abstract

The internet has been used widely in all aspects of life. The Interference of internet connections can produce a significant impact. Therefore, the role of the Network Intrusion Detection System (IDS) to detect cyber attacks is very important. A suspicious connection needs to be blocked immediately before performing anything further. The performance of an IDS depends on the algorithm and the training data used. Irrelevant features in training data can decrease the detection performance and accuracy of IDS. This research will observe the impact of using feature selection on the Intrusion Detection System. The Information Gain, Gain Ration, Chi-squared, and ReliefF selection method would be examined in J48, Random Forest, Naive Bayes, and KNN algorithm to show the effect. The results show that feature selection can enhance the performance of IDS significantly, although it makes a slight reduction inaccuracy.