The SVM intrusion detection problem based on nonlinear projection and penalty function

摘要

Based on the idea of a linear projection pursuit, we propose a method of nonlinear projection. The nonlinear projection method will reduce the high dimensional data into low-dimensional space. For low-dimensional data projection thus obtained with a penalty function reuse nonlinear support vector model and implement intrusion detection data. Finally, we use the KDD99 data set to illustrate the model's effectiveness. Verified by calculation, the effect is more ideal.