Use of Machine Learning and Deep Learning in Intrusion Detection for IoT

The ever-increasing use of IoT devices has presented new security threats and, thus, requires IDS to protect interconnected IoT networks. This paper identifies the use of ML and DL as a new effective way of improving protection against cyber threats in IoT networks. This review aims to review the most up-to-date research on ML and DL techniques for IoT-based IDS concerning responding to new threats like zero-day attacks and Distributed Denial of Service (DDoS). Thus, the review uses twenty cognate peer-reviewed studies published between 2023 and 2024 to emphasize the methodological variability, datasets, as well as the forms of performance metrics present in the field. The outcomes show that four modern frameworks, including hybrid models, federated learning, LSTMs, and convolutional architectures, perform much better than conventional methods in terms of accuracy, detection rates, and false-positive rates. For example, models such as feature selection employing new paradigms like the inclusion of new features in the suite, cost-sensitive learning, as well as multitask paradigms show better scalability and flexibility to handle imbalanced datasets as well as cyber-attacks of other unseen types. Also, the incorporation of IDS with energy-efficient protocols and fog computing more real-time capability of such systems within IoT-constrained resources network. However, a number of limitations including computational complexity, privacy issues, and lack of a proper baseline for model comparison remain major ongoing issues. Overall, this review consolidates important ideas about the advantages and weaknesses of current methods together with directions for further investigation such as the practices of federated deep learning, adaptive algorithms, and real-time anomaly detection frameworks. In conclusion, this paper establishes the importance of ML and DL in enhancing the robustness of IoT systems to the increasing ecosystem of cybersecurity threats.