WHTE: Weighted Hoeffding Tree Ensemble for Network Attack Detection at Fog-IoMT

Abstract

The fog-based attack detection systems can surpass cloud-based detection models due to their fast response and closeness to IoT devices. However, current fog-based detection systems are not lightweight to be compatible with ever-increasing IoMT big data and fog devices. To this end, a lightweight fog-based attack detection system is proposed in this study. Initially, a fog-based architecture is proposed for an IoMT system. Then the detection system is proposed which uses incremental ensemble learning, namely Weighted Hoeffding Tree Ensemble (WHTE), to detect multiple attacks in the network traffic of industrial IoMT system. The proposed model is compared to six incremental learning classifiers. Results of binary and multi-class classifications showed that the proposed system is lightweight enough to be used for the edge and fog devices in the IoMT system. The ensemble WHTE took trade-off between high accuracy and low complexity while maintained a high accuracy, low CPU time, and low memory usage.