Encrypted traffic classification is a key task in cyberspace security and network management, whose core objective is to accurately identify the application services and user behaviors corresponding to traffic by analyzing the side-channel information of network data packets. Machine learning methods have achieved remarkable progress in this field, however, facing the demand for processing massive network data and deploying models on edge devices, existing approaches still confront severe challenges. To address this, this paper proposes a dual encoding hybrid quantum convolutional neural network model named DE-HQC. The model extracts the header and payload byte sequences from network data packets respectively, performs feature extraction on the two types of sequences via amplitude encoding combined with a quantum convolutional neural network, and finally obtains the recognition result through feature fusion and a classifier. Experiments on two public encrypted traffic datasets, ISCX-VPN and USTC-TFC, demonstrate that DE-HQC achieves accuracies of 90. 55\% and 99. 45\% respectively, and its classification performance is superior to most classical methods and other quantum neural network baselines, while also exhibiting an architecturally compact design with fewer model parameters and lower computational overhead.
Zeng et al. (Sat,) studied this question.