Affiliations: [a] College of Computer Science and Engineering, Chongqing Three Gorges University, Chongqing, China
| [b] College of Electronic and Information Engineering, Chongqing Three Gorges University, Chongqing, China
| [c] Department of Computer Science, University of Rochester, New York, USA
Correspondence:
[*]
Corresponding author. Weijian Fang, College of Computer Science and Engineering, Chongqing Three Gorges University, Chongqing, China. E-mail: [email protected].
Abstract: With the increasingly serious network security situation, intrusion detection technology has become an important means to ensure network security. Therefore, it has become a consensus to introduce the theory and method of machine learning into intrusion detection, and has made good progress in this research field in recent years. In this paper, a machine learning intrusion detection system is proposed. The system uses the intrusion detection of Elman neural network and the intrusion detection of robust SVM neighbour classification to solve the above problems. Elman neural network intrusion detection uses clustering algorithm to cluster the text of the network packet, which overcomes the defect of missing the text information of the network packet. At the same time, the ability to detect abnormal behaviour between network packet sequences is improved. At the same time, robust SVM neighbour classification intrusion detection can achieve the feature space weighting of the optimal classification face host system log, eliminate the negative impact of noise data, reduce the false alarm rate of intrusion detection, and improve the detection accuracy. Under the requirement of false alarm rate of 0, the intrusion detection based on robust SVM neighbour classification can achieve 87.3% detection rate; when the false alarm rate is 2.8%, the detection rate is 100%.
