Research on Lightweight Learning in the Field of Traffic Analysis

Internet traffic analysis is a key link in the field of network management and security. However, with the rapid development of Internet traffic in recent years, people have increasingly high requirements for privacy protection, and encryption traffic has increased dramatically. Encrypted Traffic Classification (ETC) has become an important direction in network management and security research. The existing encryption traffic classification methods are mainly divided into machine learning based (ML) and deep learning based (DL). ML based methods typically require experts to manually extract traffic features, which requires classifiers to observe the entire traffic or most data packets to obtain features, making them more suitable for offline classification. At the same time, existing deep learning (DL) based methods only blindly pursue the accuracy of traffic classification while ignoring the scale and efficiency of the model. Starting from the needs of actual network traffic management and security analysis, this article lists the existing lightweight models for encrypted traffic classification in recent years. Based on existing research results, the advantages and disadvantages of each model are systematically summarized and compared from multiple aspects such as data processing, identification methods, accuracy, and performance indicators. Finally, based on the current research, combined with the development trend of the future Internet network environment and the practical problems of DL model research, this paper analyzes and prospects the research direction of the further lightweight encryption traffic classification model.