Impact of Feature Selection Algorithms on Network Intrusion Detection

This research paper discusses the importance of efficient network intrusion detection as a key element in the field of cybersecurity. Network Intrusion Detection Systems (NIDS) are crucial for spotting and stopping unwanted activity on computer networks as cyber-attacks become more sophisticated. Various feature selection algorithms have been used in the literature for network intrusion detection systems. In this work, we select three feature ranking techniques namely, Anova F-test, Mutual Information, and Chi-square test to rank the traffic features. The UNSW-NB15 dataset was used in the research to assess the effectiveness of five well-known machine learning models, namely Logistic Regression, Decision Trees, Random Forest, K-Nearest Neighbours, and Naive Bayes. In order to assess how feature selection affects algorithm accuracy, we compare the detection accuracy obtained from each of the feature ranking techniques. The study emphasizes how crucial accurate feature selection is in improving NIDS’ accuracy and lowering false positives. The results also show that each algorithm gives the highest accuracy at varying numbers of features. Overall, this research offers insightful information about the efficacy o f machine learning algorithms and feature selection techniques for network intrusion detection. The findings h ave s ignificant ramifications for enhancing computer network security and defending against cyber threats.