Development of K-Neareast Neighbor Based Model for Network Intrusion Detection Using Random Forest for Feature Selection

Abstract

Networked monitoring has emerged as one of the primary security technological advances, and network security challenges have gained increasing attention due to the extensive usage of the web. As a result, conventional signature-based and anomaly-centered intrusion detection systems are not effective enough. The efficacy and preciseness of network security measures are significantly impacted by the beginning information source's substantial dimension and volume of content. An inventive method for choosing the key attributes and reducing the overall length of information was the random forest-based feature selection, which simultaneously increased the Ture Positive Rate and decreased the False Positive Rate. Consequently, the classifier and selection of features both contribute significantly to improving network monitoring performance. On the simulated dataset, the suggested K-nearest neighbor and Random Forest perform admirably and successfully ensure the kNN classifier's progress. Simulations and experiments were conducted using the R programming language, and the results indicate that the model we developed reaches 99% accuracy. The experimental findings demonstrate the effectiveness