Machine Learning-based Intrusion Detection System for IoT using Contiki-OS

The rapid expansion of the Internet of Things (IoT) is a serious concern in terms of security due to the resource-constrained environment. Classical IDS cannot work properly in such scenarios because computation cost is too high. This paper proposed a machine learning-based IDS for Contiki-OS on IoT networks that adopts the AdaBoost algorithm and a Random Forest approach in the detection of cyberattacks such as DDoS, sinkhole, and MITM attacks. The system is trained on the Edge-IIoTset dataset with data preprocessing to realize optimal performance. The simulated results in the Cooja network simulator are complemented by validation on real IoT hardware to reveal the effectiveness of the IDS. Results in 91.08% accuracy and 93.0% precision with 92.0% recall for the ability to improve IoT security with low resource usage. Future enhancement involves integrating deep learning with enhanced attack coverage.