Minimising False Alarm Rate in Network Intrusion Detection System Model Using KNN Classifier and Chi-Square for Feature Selection

Abstract

False alarms in network intrusion detection systems (NIDS) can lead to unnecessary and costly investigations and reduce the credibility of the system. Network intrusion detection systems (NIDS) are tools that monitor network traffic and detect malicious or anomalous behaviours. However, NIDS face many challenges, such as the high dimensionality of network data, class imbalance, evolving types of attacks, and high false alarm rates. Therefore, there is a need to develop effective and efficient methods for NIDS that can achieve high accuracy and low false alarms. In this study, we propose a method for NIDS based on the use of the KNN classifier and chi-square for feature selection. The KNN classifier is a simple and robust machine-learning algorithm that assigns a class label to a new instance based on the majority vote of its k-nearest neighbours in the training data. Chi-square feature selection is a statistical technique that ranks features based on their dependency on the class label and selects only the relevant features for classification. The proposed method aims to minimise the false-alarm rate and improve the accuracy of a NIDS by reducing the dimensionality of the network data and enhancing the performance of the KNN classifier. We evaluated our method using the CSE-CIC-IDS2018 dataset and compared it to other related methods. The experimental results show that our method achieves a high accuracy of 99.9% and low false-alarm rates for both false positives and false negatives 3 and 9.

Keywords: Network intrusion detection, false alarm rate, KNN classifier, chi-square, feature selection.