An Advanced Feature Selection Approach to Improve Intrusion Detection System using Machine Learning

This research study presents an efficient threat detection model in which primary emphasis has been placed on the feature selection and classification phase The primary objective of the proposed model is to enhance the accuracy of intrusion detection while decreasing its complexity and false alarm rates. In this model, entropy-based infinite feature selection, Eigenvector centrality, and ranking feature selection algorithms are used for selecting only significant and critical features. Moreover, three classifiers, i.e., Artificial Neural Network (ANN), K-nearest neighbor (KNN), and Decision Tree (DT), are used as classifiers for detecting and categorizing potential attacks that may take place in the IoT network. KDD-CUP99 and NSL-KDD were used as benchmark databases to test the performance of the suggested technique in MATLAB. The results revealed that the proposed DT and KNN are generating the most promising results compared to the rest of the models in terms of accuracy, precision, recall, and F1 score.