Rapid advancements in the internet and networking fields have culminated in a massive increase in network size and data. As a result, several novel attacks are being created, posing challenges for network protection to detect intrusions accurately. An intrusion detection system (IDS) is one such mechanism that guards against potential network intrusions by monitoring network traffic to ensure its security, credibility, and availability. Despite tremendous efforts by researchers, IDS continues to face difficulties in enhancing detection accuracy while reducing false alarm rates and detecting novel intrusions. Recently, machine learning (ML) and deep learning (DL)-based IDS have been implemented as possible solutions for efficiently detecting network intrusions. This article first defines IDS and then includes a taxonomy focused on prominent ML strategies used in the design of network-based IDS (NIDS) structures. A thorough analysis of recent NIDS-based publications is presented, with an emphasis on the strengths and shortcomings of the proposed solutions.