Online Encrypted Traffic Classification Based on Lightweight Neural Networks *
Yulei Wu, Jingguo Ge, Tong Li · 2022
Due to the increased awareness of privacy protection, the surge in the volume of encrypted traffic challenges the efficiency of network management systems. Besides, the nontransparency of encrypted traffic leads to high-computational overheads in making traffic classification, which makes efficient network management even harder. Existing traffic classification approaches sacrifice the efficiency to obtain high-precision classification results, which are no longer suitable for scenarios with a large volume of encrypted traffic. To reduce the computational overheads, in this chapter, a lightweight and online approach for traffic classification is introduced, which adopts the multihead attention mechanism and the convolutional networks. Due to the one-step interaction of all packets and the parallel computing, the multihead attention mechanism can significantly reduce the number of model parameters and the model running time. In addition, the effectiveness and efficiency of convolutional networks are proved in traffic classification. Comparisons with the existing state-of-the-art models on three typical datasets demonstrate that the proposed model has higher accuracy and running efficiency.