A Naive Feature Selection Method and Its Application in Network Intrusion Detection

Network intrusion detection system needs to handle huge data selected from network environments which usually contain lots of irrelevant or redundant features. It makes intrusion detection with high resource consumption, as well as results in poor performance of real-time processing and intrusion detection rate. Without loss of generality, feature selection can effectively improve the classification model performance, study on the feature selection-based intrusion detection method is therefore very necessary. This paper proposes a simple and quick inconsistency-based feature selection method. Data inconsistency is firstly employed to find the optimal features, and the sequential forward search is then utilized to facilitate the selection of subset features. The tests on KDD99 benchmark data show that the proposed feature selection method can directly eliminate irrelevant and redundant features, without degenerating the classification performance. Furthermore, due to experiments, the intrusion detection performance using the proposed method is also a little advantageous than that with the general CFS method.