Performance Analysis of Machine Learning Models in Network Intrusion Detection Systems

Networks plays a vital role in modern society and cyber security has become a critical area that needs to be acknowledged. A crucial part of cybersecurity frameworks intended to track network traffic and spot suspicious activity, and malicious activity is an intrusion detection system(IDS). Integrating machine learning (ML) algorithms into IDS greatly improves the system’s ability to study complex network traffic and accurately identify abnormal patterns. In addition, the incorporation of algorithms reduces the complexity of large data sets and the training time of models. In this paper, we propose the application of three algorithms Decision Tree(J48), SVM, and Naïve Bayes on the NSL-KDD dataset. These algorithms are applied after preprocessing dataset, which includes feature selection and class imbalance resolution. After applying the aforementioned algorithms, we assessed their performance based on metrics such as confusion matrix, accuracy. The results of the environment clearly indicate which algorithm works best with the data set and achieves high accuracy with minimal time required to build the model.