Attribute Reduction Method Applied to IDS

Abstract

In this paper, we apply a new linear correlation attribute reduction algorithm to feature selection. The algorithm is valuable when the features are marginally unrelated but jointly related to the response variable. A new technique is introduced to remove redundant attributes and it is effective to reduce the false selection rate in the feature selection stage. We train and test the new algorithm on KDD1999 data set, and compare the experiment results to illustrate the methodology.