Encrypted Network Traffic Classification: A data driven approach

• Published in: 2020 IEEE Intl Conf on Parallel & Distributed Processing with Applications, Big Data & Cloud Computing, Sustainable Computing & Communications, Social Computing & Networking (ISPA/BDCloud/SocialCom/SustainCom) pp. 706 - 712
• Main Authors: Zhang, Zhongkai, Liu, Lei, Lu, Xudong, Yan, Zhongmin, Li, Hui
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.12.2020
• Subjects: Convolutional neural networks; Cryptography; encrypted traffic classification; Feature extraction; Internet; one-dimensional convolutional neural networks; Telecommunication traffic; Training; two-stage
• DOI: 10.1109/ISPA-BDCloud-SocialCom-SustainCom51426.2020.00114

With the growing number of the network based applications, the quality and usability of various applications require distinguished requirements in light of network performance. The Internet is a vast network of independently-managed networks. The traffic is heterogeneous and consists of flows from multiple applications. For the sake of Internet governance, identifying who is who has become an impressing demand. The purpose of Internet governance is to properly distinguish the demands across the limited recourse. Traffic classification is the very beginning that helps identify who and how much the application requires. Hence, the classification and identification of encrypted traffic has drawn much research attention. Compare to the traditional methods such as port-based methods and payload-based methods, this study establishes a two-stage traffic classification framework based on convolutional neural network. Experimental results suggest that the developed method has a good degree of traffic classification subject to fewer bytes in the packet.