Traffic Identification and Classification Method for Encrypted Tunnels based on Heterogeneous Dual-Stream Network

• 1、School of Computer Science (National Pilot Software Engineering School),Beijing University of posts and telecommunications,Beijing 100876

Abstract：Encryption traffic recognition is crucial for protecting network security, improving network management efficiency, and ensuring user privacy. The traditional deep learning model\'s encrypted traffic recognition methods have problems such as poor performance and weak generalization ability. To address the above issues, this paper proposes an encrypted traffic identification method based on the Heterogeneous Two Stream Network (DTSN) model. This article generates a traffic graph through the payload sequence of traffic, extracts its time feature vector using residual networks, and captures statistical features using packet information and convolutional networks. Subsequently, using feature fusion technology to achieve TLS tunnel traffic recognition, in order to improve the accuracy of recognition. Experiments were conducted on the dataset captured by oneself, and the results showed that for traffic recognition problems, the accuracy of the DTSN model can reach 99.6%, which is 4.5% higher than traditional CNN models. The above experimental results demonstrate that the recognition scheme proposed in this article has high accuracy.

Keywords： Model Identification Dual-stream Network Tunnel Technology