Anomaly-based Intrusion Detection Approach for IoT Networks Using Machine Learning

The proliferation of the Internet of Things (IoT) devices in smart environments such as smart cities or smart home facilitate communication between various objects. Nevertheless, this technological advancement comes with security challenges of IoT devices. Thus, current attacks targeting IoT networks have become motivating factors in implementing security mechanisms. Such attacks come in the form of intrusion or anomalies. Anomaly detection mechanisms have been implemented to prevent confidential resources from malevolent users. Therefore, this paper presents a new anomaly-based approach for IoT networks which is implemented with a hybrid feature selection engine that only selects most relevant features; and the Random Forest algorithm which classifies each traffic as normal or anomalous. The performance was evaluated using IoTID20, one of the latest anomaly detection datasets collected in the IoT Environment. The experimental results show that the proposed method achieves relatively high accuracy while detecting DoS (99.95%), MITM (99.97%), Scanning (99.96%) attacks.