Intrusion Detection System Based on Improved SVM Incremental Learning
Hongle Du, Shaohua Teng, Mei Yang, Qingfang Zhu · 2009
When collecting network connection information, we can not obtain a complete data set at once, which result in SVM training insufficiently and high error rate of prediction. To solve this problem, this paper proposes a new method that combines support vector machine with clustering algorithm, based on analyzing the relation between boundary support vectors and KKT condition. In the method, firstly, presents incremental support vector machine learning algorithm based on clustering, and describes the running process of the algorithm detailedly; then give the intrusion detection model based on incremental SVM learning; finally, the performance of the model is tested by computer simulation with KDD CUP1999 data set. The experimental results show it has higher detection accuracy rate and improves the speed of SVM training and classification, as keeping the generalization ability of incremental SVM learning algorithm.