Wrapper Feature Selection Based on Lightning Attachment Procedure Optimization and Support Vector Machine for Intrusion Detection

As the internet becoming omnipresent, a large number of attacks exist from the inside of the network. The intrusion detection system, one of the most effective way to monitor the network for defending inner attacks, which is gaining more and more attention. However, in the process of the network intrusion detection, feature redundancy might reduce the accuracy of classification or clustering, increase the time and space complexity and bring down the learning performance and efficiency of the algorithm as well. In the paper, a wrapper feature selection method based on lightning attachment procedure optimization algorithm (LAPO) and support vector machine (SVM) for intrusion detection are proposed. LAPO is a newly proposed nature-inspired algorithm that has robust searchability. For evaluating the performance of the proposed method, the popular KDD Cup 99 dataset is employed. Compared with genetic and particle swarm optimization algorithm, experimental result shows the proposed approach presents a better efficiency and accuracy in searching for the optimal feature subset.