Performance Evaluation of Ensemble Machine Learning Techniques for IoT Edge Device Security

Machine learning based Intrusion detection (ML-IDS) has been long enforced for the protection of the IoT against malicious attacks. Researchers focused on improving ensemble intrusion detection methods in order to boost the efficiency of attack detection on resource limited IoT edge devices. However they confined their evaluation platforms to cloud or standalone computing facilities rather than the targeted physical devices. In this paper, we complement our work with feasibility study of managing ensemble-based ML-IDS combiner models on a resourceful physical edge device, and demonstrate the implied limitations of deploying such models in the IoT without cloud support. We also elaborate on the potential of cloud support to support distributed ensemble techniques, resulting in super container models that harness limited resources of IoT edge devices.