A novel encrypted traffic classification method with attention mechanism and Haar wavelet transform

As the information age rapidly evolves, the encrypted traffic in the network exhibits an exponential growth trend. Encrypted traffic classification technology has become the key to ensure information security and optimize network operation. However, the traditional traffic classification methods based on port number and payload cannot meet the increasing user privacy protection requirements and the demand for fast real-time classification. Meanwhile, current deep learning-based traffic classification approaches suffer from issues like inadequate data processing, limited feature extraction, and imbalanced sample category distribution, leading to poor classification performance. To solve these problems, we propose a residual network traffic classification model based on Convolutional Block Attention Module (CBAM) and Haar Wavelet Downsampling (HWD), aiming to fully exploit the intrinsic features between traffic bytes. And we validate the accuracy and effectiveness of the proposed method using the publicly available ISCX VPN-nonVPN dataset. The experimental results demonstrate that our model performs better in terms of accuracy, precision, recall and F1-score compared with existing methods, which has important theoretical significance and potential for practical application.