A Deep Learning-Based Encrypted VPN Traffic Classification Method Using Packet Block Image

Network traffic classification has great significance for network security, management and other fields. However, in recent years, the use of VPN TLS encryption had presented with new challenges. Due to performances deep learning image recognition, many solutions have focused on learning-based method achieved positive results. A based is provided this paper, where concept Packet Block proposed, which aggregation continuous packets same direction. The features are extracted from traffic, then transformed into images. Finally, convolutional neural networks used identify application type traffic. experiment conducted using captured OpenVPN dataset public ISCX-Tor dataset. results shows that accuracy 97.20% 93.31% dataset, higher than state-of-the-art methods. This suggests our approach ability meet challenges encryption.