IGRF-RFE: a hybrid feature selection method for MLP-based network intrusion detection on UNSW-NB15 dataset

Abstract The effectiveness of machine learning models can be significantly averse to redundant and irrelevant features present in the large dataset which cause drastic performance degradation. This paper proposes IGRF-RFE: a hybrid feature selection method tasked for multi-class network anomalies using multilayer perceptron (MLP) network. IGRF-RFE exploits qualities both filter its speed wrapper relevance search. In first phase our approach, we use combination two methods, information gain (IG) random forest (RF) respectively, reduce subset search space. By combining these influence less important but with high-frequency values selected by IG is more effectively managed RF resulting relevant included second learning-based that provides recursive elimination (RFE) further dimensions while taking into account similar features. Our experimental results obtained based on UNSW-NB15 confirmed proposed improve accuracy anomaly detection as it select reducing show reduced from 42 23 multi-classification MLP improved 82.25% 84.24%.