Although the Internet of Things (IoT) devices simplify and automate everyday tasks, they also introduce a tremendous amount of security flaws. The current insufficient security measures for smart device protection make IoT devices a potential victim of breaking into a secure infrastructure. This research proposes an on-the-fly intrusion detection system (IDS) that applies machine learning (ML) to detect network-based cyber-attacks on IoT networks. A lightweight ML model is trained on network traffic to defer benign packets from normal ones. The goal is to demonstrate that lightweight machine learning models such as decision trees (in contrast with deep neural networks) are applicable for intrusion detection achieving high accuracy. As this model is lightweight, it could be easily employed in IoT networks to classify packets on-the-fly, after training and evaluation. We compare our lightweight model with a more complex one and demonstrate that it could be as accurate.
