Network Anomaly Detection Based on Cooperative Semi-Supervised Support Vector Machine
Lin Zhang, Hongle Du, Yan Zhang · 2019
Network behavior data is imbalanced data and the classification hyper plane will shift in imbalanced data. Combined with cost sensitivity, voting mechanism and Transductive Support Vector Machine, the Cost Sensitivity Cooperative Semi-supervised Support Vector Machine is proposed to resolve this problem. In this algorithm, data set is divided by clustering, and the cost sensitivity is computed according to the class sample distribution. Then the sample is tagged according to the voting results of every sub-classifier. Finally, all sub-classifiers are integrated to obtain the final classifier and the classification performance of the final classifier is improved. This algorithm has better generalization performance for imbalanced data and can improve the recognition rate of new intrusions. Finally, experiment results with KDDCUP99 data set show the algorithm can improve overall classification performance and improve the detection accuracy for Unknown attacks.