Deep Learning Based Malware Traffic Classification for Power Internet of Things Network Security

Abstract

The detection and classification of Malware traffic is an essential building block for the network security of power Internet of Things. Recent studies show that deep learning models can be applied for accurate network traffic classification, but most of the existing works choose to convert the traffic flows into images and turn the problem into image classification. In this work, we propose a novel deep learning-based malware traffic classification approach for power Internet of things network, which represents the traffic flows in the form of fixed-size byte sequences and builds 1D-CNN models for malware traffic detection and classification. For evaluation, we compare the proposed approach to the existing deep learning-based model by using a publicly available malware traffic data set. The results show that the 1D-CNN models trained by the proposed approach outperform the existing models throughout different scenarios in the experiments.