关键词: 入侵检测, 特征选择, 特征关联性, 改进粒子群算法

Abstract: In order to improve the detection accuracy of network intrusion, this paper proposes a novel network intrusion detection model （KNN-IPSO） based on K Nearest Neighbor and Improved Particle Swarm Optimization algorithm to select the features. Firstly, KNN algorithm is used to eliminate redundant features and remove the redundant features of network data and the selected features are taken as the initial solution of PSO, secondly, the optimal features are obtained by PSO which inertia weight is adjusted adaptively and chaotic system is intruded to help the swarms jump out the local optimal solutions, finally the KDD CUP 99 data sets are used to test the performance of KNN-IPSO. The results show that the proposed model can eliminate the redundant features and reduce the input dimensions of classifier. It can improve the network intrusion detection accuracy and detection speed.

Key words: intrusion detection, feature selection, feature relevance, improved particle swarm optimization algorithm