Abstract
In this work, homogeneous ensemble techniques, namely bagging and boosting, were employed for intrusion detection to determine the intrusive activities in network by monitoring the network traffic. Simultaneously, model diversity was enhanced as numerous algorithms were taken into account, thereby leading to an increase in the detection rate. Several classifiers (i.e., SVM, KNN, RF, ETC, and MLP) were used in case of bagging approach. Likewise, tree-based classifiers have been employed for boosting. The proposed model was tested on NSL-KDD dataset that was initially subjected to preprocessing. Accordingly, the 10 most significant features were identified using decision tree and recursive feature elimination method. Furthermore, the dataset was divided into five subsets, each one being subjected to training, and the final results were obtained based on majority voting. Experimental results proved that the model was effective for detecting intrusive activities. Bagged ETC and boosted RF outperformed all the other classifiers with an accuracy of 99.123% and 99.309%, respectively.