运用核Fisher鉴别分析和MPM分类器的入侵检测

Intrusion Detection Based on Kernel Fisher Discriminant Analysis and Minimax Probability Machine Classifier

为了提高分类器的正确率和减少训练时间,将特征提取技术与分类算法结合,提出了一种基于核Fisher鉴别分析和最小极大概率机算法的入侵检测算法。利用核Fisher鉴别分析技术提取关键特征,运用最小极大概率机对提取特征后的数据进行分类,采用离线数据集KDDCUP99进行实验。实验结果表明,该算法是可行和有效的,使分类性能和训练时间都得到了提高。

To improve the performance of Minimax Probability Machine （MPM） in the detection rate and the training time, Intrusion Detection Based on Kernel Fisher Discriminant Analysis and Minimax Probability Machine Classifier （KFDA-MPM） algorithm is proposed which combines the feature extraction technology and classification algorithm. In this method, the KFDA is used to extract the optimal feature set and then the MPM is adopted to classify the optimization data. Results of the experiment using the Knowledge Discovery and Data Mining Cup 1999 （KDDCUP99） datasets indicate the effectiveness of the algorithm.