Context adaptive ensemble classification mechanism with multi‐criteria decision making for network intrusion detection

Abstract

SummaryNowadays, intrusion detection systems (IDSs) enabled with computational intelligence for electing the suitable classifiers to learn the patterns of various types of network attacks are under study. Selection mechanism of single or ensemble classifiers based on performance and ranking is to be enabled in the IDS framework to make accurate pattern predictions. A new anomaly based network intrusion detection system framework is proposed using context adaptive classification through the technique for order of preference by similarity to ideal solution (TOPSIS) ranking mechanism with hierarchy based chi‐square and bat algorithms for feature selection. Parameters like accuracy, false positive rate (FPR), and classification model building time are explored for choosing the best and situation aware classifiers using TOPSIS. For experimentation, NSL‐KDD and UNSW‐NB15 datasets are used. The proposed system opts decision tree (DT) algorithm for NSL‐KDD and produces 98.77% accuracy, 0.03% FPR with 8 features. Ensemble learner is selected for the UNSW‐NB15 dataset using DT and support vector machine classifiers with 9 features. Combined classifier predicts 89.43% accuracy and 3.215% FPR. Experimentation on the proposed methodology with the state of the art approaches produces promising results.