An Improved Binary Whale Optimization Algorithm for Feature Selection of Network Intrusion Detection

Abstract

As for network intrusion detection, the time and space consumption of the detection algorithm increases with the feature dimension. The irrelevant or redundant features may cause the detection accuracy to decrease. In order to improve the timeliness and reliability of intrusion detection, feature selection is one of the most effective methods proposed currently. Considering that Whale Optimization Algorithm (WOA) has good classification performance and dimensionality reduction in reducing feature redundancy, it can be applied to feature selection of network intrusion detection. The WOA converges slowly in the search process and easily falls into a local optimum in the updating mechanism, which restricts the classification performance and dimensionality reduction of the algorithm. Therefore, this paper adopts the nonlinear convergence factor strategy in the search process and introduces the Particle Swarm Optimization (PSO) strategy in the updating mechanism. An Improved Binary Whale Optimization Algorithm (IBWOA) is then proposed in order to achieve a better dimension reduction on the premise of ensuring accuracy for feature selection of intrusion detection. In the experimental stage, multiple data sets in the UCI database for machine learning were used for verification by comparing with other feature selection methods. Finally, the KDD CUP 99 data set was used for verification of feature selection in intrusion detection. The experimental results show that, compared with Genetic Algorithm (GA), PSO and WOA, the IBWOA improves the classification accuracy and dimensionality reduction in feature selection of intrusion detection.