期刊
出版社
IEEE
DOI: 10.1109/ICTC52510.2021.9620746
关键词
Traffic Classification; Encrypted Traffic; Machine Learning; Deep Learning; Traffic Identification
资金
- Priority Research Centers Program through NRF - MEST [2018R1A6A1A03024003]
- Grand Information Technology Research Center support program by MSIT, Korea [IITP-2021-2020-001612]
This survey paper introduces the emerging research and general framework for deep learning-based methods for traffic classification. Deep learning methods are shown to provide better accuracy and automatic feature learning capabilities compared to traditional machine learning methods.
Network traffic classification plays an important role in various network functions such as network security issues and network management. In addition to port-based and payload-based approaches, the classical machine learning approaches have been studied for past decades, but there are some limitations, namely time-consuming, frequent features updates, and the accuracy has decreased due to the rise of internet traffic, especially encrypted traffic. Deep learning comes with the ability of automatic feature learning, some studies try to apply it and reported better accuracy. This survey paper introduces the emerging research and general framework for deep learning-based methods for traffic classification. We present commonly used deep learning methods and their application in traffic classification tasks.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据