Feature Importance Ranking for Increasing Performance of Intrusion Detection System

The performance of the Intrusion Detection System (IDS) depends on the quality of the model generated in the training process. An appropriate process positively affects not only the performance but also computational time for detecting intrusions. Reliable training data can be obtained by preprocessing the dataset, which can be feature extraction, reduction, and transformation. Generally, feature selection has become the main problem. In this research, we work on that issue by developing a new method based on Feature Importance Ranking Classification. We propose to reduce the size of the dimension by combining Feature Importance Ranking to calculate the importance of each feature and Recursive Features Elimination (RFE). The results of the experiment show that the proposed method raises the performance over the existing methods. It can be proven by evaluating some metrics: accuracy, sensitivity, specificity, and false alarm rate.