A Multi-Distance Ensemble and Feature Clustering Based Feature Selection Approach for Network Intrusion Detection
Guanghua Fu, Bencheng Li, Yongsheng Yang, Qingjuan Wei · 2022
Due to the rapid development of IoT technology, a tremendous amount of terminals have been connected with each other. Meanwhile, the means and methods of network intrusion have become more complex. Therefore, the role of the Network Intrusion Detection System (IDS) based on machine learning is very important to detect the attacks. One of the biggest challenges in IDS is the high-dimensional and imbalanced data which increase the cost of machine learning. To solve this problem, this paper proposes a feature selection method based on multi-distance ensemble and feature clustering. Four different distance metrics firstly combined with ReliefF algorithm to produce an ensemble feature subset. Then Fuzzy C-Means clustering algorithm is utilized to cluster this subset. The final features would be selected from the different clusters with less feature redundancy. The proposed method is next tested with Support Vector Machine and K-Nearest Neighbors classifiers over NSL-KDD dataset. The results show that it yields better performance than other feature selection algorithms in terms of accuracy and F-measure.