A Comprehensive Review: Intrusion Detection System Using Machine Learning in Internet of Things

The Internet of Things (IoT) refers to a wide range of internet-connected devices used for data collecting, transfer, and dissemination. The increasing use of these devices causes important privacy and security issues. More attention is being given to internet security to protect a large number of vulnerable devices from harmful online activities. The increasing complexity of these threats has led to the need for an Intrusion Detection System (IDS) designed to detect and mitigate these security threats. However, Machine Learning (ML) refers to develop algorithms and statistical models that enable computers to autonomously select a particular task by identifying patterns in large datasets without explicit programming, such as supervised, unsupervised, and semi-supervised learning. This survey analyzes multiple implemented ML strategies in IDS for IoT, which adapt their techniques to the new needs of IoT environments and data security. The best result was achieved by Random Forest and XG-Boost for F1-score and precision nearing 100% with high accuracy rates up to 99.6% on the UNSW-NB15 dataset. Implementing ML with IDS in IoT networks effectively enhances security by identifying threats, addresses the complex and evolving nature of cyber threats, and offers improved detection accuracy and adaptability in dynamic IoT environments, therefore supporting a more resilient and secure IoT.