Improving Classification Accuracy of Intrusion Detection System using Feature Subset Selection

Intrusion detection system (IDS) research field has grown tremendously in the past decade. Improving the detection rate of user to root (U2R) attack class is an open research problem. Current IDS uses all data features to detect intrusions. Some of the features may be redundant to the detection process. The purpose of this empirical study is to identify the important features to improve the detection rate and reduce the false detection rate. The investigated feature subset selection techniques improve the overall accuracy, detection rate of U2R attack class and also reduce the computational cost. The empirical results have shown a noticeable improvement in detection rate of U2R attack class with feature subset selection techniques.