Cybersecurity has emerged as an essential field of study due to the ubiquitous nature of networks in contemporary society. A network intrusion detection system (NID) is a critical component of network security. Despite significant advancements over the years, existing intrusion detection systems still need improvement in terms of detecting new threats, reducing false positives, and increasing detection accuracy. They strengthen the network’s defenses against the expanding variety of threats. Intrusion detection benchmark datasets aim to realistically represent network traffic by having more examples of benign traffic than malicious ones. This creates inconsistencies in the NIDS’s training data and makes it harder for it to recognize specific forms of data imbalances. We study benchmark datasets, metrics, and intrusion detection systems that strongly depend on machine learning (ML) and deep learning (DL) models. This research helps academicians and research scholars to get a brief analysis report on the class-imbalance handling models and intrusion detection models that help to design innovative solutions for handling these issues.