The importance of intrusion detection is a foremost issue in the field of cyber security. Vulnerabilities in network security have surged due to increased network traffic and the internet’s rapid growth. Cybercriminals exploit flaws in network design to disrupt computer systems or steal sensitive information. Timely intrusion detection and protection of network resources from them are the main purposes of the Intrusion Detection System (IDS). The implementation of IDS uses various machine learning methods to identify inconsistencies in network traffic and address network security concerns. A variety of datasets, including attack patterns, train these IDSs. The IDS model struggles to identify some minor attacks due to missing patterns in the training set, resulting in a reduced detection rate for rare attacks. IDS faces a primary challenge in enhancing the detection rate of minor attacks in the dataset, a problem that the literature has not yet addressed. We utilize different machine learning algorithms to detect attacks, achieving an accuracy of above 99%, but they fail to detect rare attacks. We implemented the Synthetic Minority Oversampling Technique (SMOTE) to address imbalanced classes and enhance the accuracy of classification models. The Random Forest classifier, along with the SMOTE, achieved 99.48% accuracy when 28 features were considered using the low variance filter method.