A New Data-Based Method for Intrusion Feature Selection

摘要

The data processed by the intrusion detection system contain large numbers of redundancy and noise features,which result in long training time and bad performance of the detecting ability.This paper gives a new data-based method for the SVM feature selection,which embeds the feature selection to the same studying framework of SVM directly,and onlyselects the features by the given training samples.It is also proved that the algorithm is effective in the datasel of KDD CUP 99.