Study on Genetic Algorithm Optimization for Support Vector Machine in Network Intrusion Detection

Abstract

Abstract This paper studies on methods taken in solving the existing network disorders in network intrusion detection. Traditional parameter-optimized Support Vector Machine (SVM) may easily generate improper parameter-selections, and may further lead to a low accuracy in network intrusion detection. In order to overcome such problems, so well as to ensure the network security, this paper tries to put forward with a Genetic Algorithm optimized Support Vector Machine in network intrusion detection. For this purpose, some procedures have to be taken. Firstly the network intrusion data should be normalized and simplified for inputs, and then to obtain optimal parameters through the parameteroptimization for SVM with a Genetic Algorithm; Finally to get network intrusion results through detecting the network data normalized by an optimal Support Vector Machine (SVM) model. The simulating results show that, compared with the traditional network intrusion detection methods, the optimized Support Vector Machine (SVM) model with a Genetic Algorithm has a high accuracy, and a low rate of losing or false alarms, and is proved to be effective in network intrusion detection.