Intrusion detection systems (IDS) are a key component in protecting computer systems and networks from cyber threats. In the past few years, the application of algorithms based on machine learning (ML) and deep learning (DL) methodologies for IDS has garnered a significant amount of interest due to their capacity to analyse enormous amounts of data and identify complex patterns of harmful activity. This ability has contributed to the rise in popularity of these techniques. This paper reviews the current state of ML and DL-based IDS, including both traditional and emerging techniques. It also discusses the key challenges and opportunities in this area, including the need for diverse and robust training data, the balance between accuracy and efficiency, and the integration of IDS with other security measures. Additionally, this paper outlines future directions for research on ML and DL-based IDS, including the potential for real-time and adaptive learning, the incorporation of domain knowledge, and the use of interpretability techniques to improve transparency and accountability.