Analysis of Ensemble Classifiers with Feature Selection for an Effective Intrusion Detection Model

Abstract

Intrusion detection system (IDS) is constructed primarily to detect and block malicious attacks on a network. Several artificial intelligence techniques and genetic algorithms are utilized to construct IDS. In this paper, a model for intrusion detection system is developed using a correlation-based feature selection system integrated with bagging. This model focuses only on detection of malicious actions on a network. Correlation-based feature selection (CFS) is a feature selection technique which is based on test theory and developed beforehand by using real and artificial problems. Comparatively, it is faster and significantly improves the performance of the learning algorithm. Bagging is a machine learning algorithm which is a variation of decision tree algorithm. This paper presents the experimental results by utilizing the Kyoto intrusion detection dataset and the results show that CFS with bagging performs better in comparison with other supervised classifiers.KeywordsAdaBoostAdapativeBaggingBoostingClassifierFeature selectionIntrusionKNN