This research paper investigates the creation and assessment of a Network Intrusion Detection System (NIDS) that utilises machine learning techniques to bolster cybersecurity through improved identification of potential network breaches. We begin with an overview of NIDS architecture, fundamental concepts in network intrusion detection, and a comprehensive literature review that categorizes various types of network attacks. An examination of recent research highlights the essential performance indicators vital for Intrusion Detection Systems (IDS). These metrics encompass accuracy, precision, recall, true negative rate, false alarm rate (FAR), F-measure, and ROC curves. In our study, we assess the performance of diverse machine learning techniques for Network Intrusion Detection Systems (NIDS). The algorithms under examination include logistic regression, decision tree classifier, SKLearn’s gradient boosting, Naive Bayes, random forest classifier, and the XGBoost gradient boosting model. Each model was trained and evaluated to determine its effectiveness, with key findings focused on training and test scores to assess generalization capabilities and accuracy. Notably, the XGBoost model demonstrated the highest accuracy, achieving a training score of 99.97% and a test score of 99.65%, with the random forest classifier and decision tree models performing closely behind. Logistic regression and Naive Bayes, while slightly lower in accuracy, offered stable generalization with minimal overfitting. The findings of this investigation underscore both the advantages and drawbacks of various algorithms, with combined approaches like XGBoost and random forest showing considerable promise for highly accurate detection of intrusions in practical network settings.