The widespread adoption of computers and the internet over recent decades has significantly increased reliance on online systems. As internet usage grows, safeguarding information has become a critical focus within cybersecurity. A robust Intrusion Detection System (IDS) is essential for protecting networks and systems from unauthorized access and cyberattacks. However, conventional IDS solutions often struggle with limited real-time responsiveness and insufficient detection accuracy, posing challenges to effective network security. Our research presents the development of an IDS using machine learning algorithms to enhance detection accuracy and minimize false positives. Experiments were conducted using the NSL-KDD and InSDN datasets, which provide a diverse range of network traffic scenarios and attack patterns. Multiple classifiers, including XGBoost, K-Nearest Neighbors (KNN), Naive Bayes, Ada Boost, and Bagging, were evaluated across different feature sets to assess their performance. Feature selection methods, such as SelectKBest, were applied to optimize the models by reducing dimensionality while maintaining high accuracy. The results demonstrate that ensemble models, particularly XGBoost, achieve over 99% accuracy with better precision and recall. Key metrics such as the F1-score, Cohen’s Kappa, and Matthew’s Correlation Coefficient validate the robustness of the models across both datasets. The proposed IDS framework offers a scalable, proactive solution for network monitoring capable of addressing evolving security threats.