Intrusion Detection System Utilizing Machine Learning Classifier Algorithms and Linear Discriminative Analysis

The signature-based intrusion detection systems (IDS) function by training themselves to recognize normal network behaviour and raising alarms when they detect any unusual patterns beyond their established boundaries. However, these systems often suffer from a high false positive rate due to the challenges in precisely defining what constitutes normal and abnormal network traffic. In this study, we employ statistical techniques namely Pearson correlation coefficient (PCC) and linear discriminant analysis (LDA) to develop novel signature-based IDS models. We subsequently assess the performance of these models using the well-established NSL-KDD dataset and compare their effectiveness using Naive Bayes, decision tree, and support vector machine (SVM) machine learning classifiers. Our experimental findings demonstrate that LDA-based IDS model exhibits comparable or even better accuracy performance and detection rate in contrast to their counterparts.