An IDS is an essential component of network security that detects unauthorized access and malicious activity in network traffic. The traditional signature-based IDS cannot detect zero-day or unknown attacks; thus, machine learning algorithms are used to improve detection based on the analysis of traffic patterns. This paper compares various famous machine learning algorithms: decision trees, SVM, KNN, random forest, and Naïve Bayes within the context of IDS. We test all algorithms with the datasets KDD’99 and NSL-KDD using accuracy, detection rate, false positive rate, and computational efficiency parameters. The results will determine practical suitability in implementing IDS and direct further research.