Network Intrusion Detection System Using Ensemble Learning Approaches

Abstract

To mitigate modern network intruders in a rapidly growing and fast pattern changing network traffic data, single classifier is not sufficient. In this study Chi-Square feature selection technique is used to select the most important features of network traffic data, then AdaBoost, Random Forest (RF), and XGBoost ensemble classifiers were used to classify data based on binary-classes and multi-classes. The aim of this study is to improve detection rate accuracy for every individual attack types and all types of attacks, which will help us to identify attacks and particular category of attacks. The proposed method is evaluated using k-fold cross validation, and the experimental results of all the three classifiers with and without feature selection are compared together. We used two different datasets in our experiments to evaluate the model performance. The used datasets are NSL-KDD and UNSW-NB15.