An efficient intrusion detection technique based on support vector machine and improved binary gravitational search algorithm

'Curse of Dimensionality' and the trade-off between high detection rate and less false alarm rate make the design of an efficient and robust Intrusion Detection System, an open research challenge. In this way, we present Hyper Clique-Improved Binary Gravitational Search Algorithm based Support Vector Machine (HC-IBGSA SVM), an efficient and adaptive intrusion detection technique to improve the performance of SVM in terms of detection rate and false alarm rate. HC-IBGSA SVM employs hyper clique property of hypergraph, novel mutation operator, and Newton-Raphson inspired position update function to fasten the search for an optimal solution and to prevent premature convergence. Further, HC-IBGSA uses a weighted objective function to maintain the trade-off between maximizing detection rate and minimizing the false alarm rate and the optimal number of features. The experimental evaluations were carried out using two benchmark intrusion datasets, namely NSL-KDD CUP dataset and UNSW-NB15 dataset under two scenarios (1) SVM trained with all features, and (2) SVM trained with the optimal feature subset and model parameters obtained from HC-IBGSA in terms of various quality metrics, stability analysis and statistical test.