Minimax Probability Machine Classifier with Feature Extraction by Kernel Pca for Intrusion Detection

In this paper, we synthetically applied kernel principal component analysis method and minimax probability machine to the intrusion detection. The intrusion features were extracted by kernel principal component analysis (KPCA) and followed by minimax probability machine (MPM) for classification. The dataset kddcup99 is our experiment data, our algorithm is compared with the algorithm which applies the support vector machine and minimax probability machine without application of feature extraction to intrusion detection. The experiment result has shown that the minimax probability machine classifier with feature extraction by KPCA can greatly reduce the training time and will not degrade the classifiers´ performance.