基于机器学习的网络流量分类综述  

Overview of Network Traffic Classification Based on Machine Learning

作　　者：于治平 刘彩霞[1] 刘树新 李星[1] 王亚辉 YU Zhiping;LIU Caixia;LIU Shuxin;LI Xing;WANG Yahui(Information Engineering University,Zhengzhou 450001,China)

机构地区：[1]信息工程大学,河南郑州450001

出　　处：《信息工程大学学报》2023年第4期447-453,483,共8页Journal of Information Engineering University

摘　　要：网络流量分类对于网络优化、网络安全预警、用户个性化服务等具有重要意义。随着通信和信息技术的发展,传统的基于端口以及深度包检测的分类方法由于私有协议的广泛应用已不能满足需求。基于机器学习的分类算法被应用于流量分类研究,但加密技术为流量分类的特征提取带来一定难度。首先总结了网络流量分类的基本流程;其次分析了分类粒度及其应用场景,并对目前流量分类的主要技术按照监督学习、半监督学习、无监督学习进行了分类研究;最后对网络流量分类技术的发展趋势及面临挑战做了展望,为网络流量分类研究提供一定的参考。Network traffic classification is of great significance for network optimization,network security early warning,user personalized service and so on.With the development of communication and information technology,the traditional classification methods based on port and deep packet inspection cannot meet the needs because of the wide application of private protocols.The classification algorithm based on machine learning is applied to the research of traffic classification,but the encryption technology brings some difficulties to the feature extraction of traffic classification.First,the basic process of network traffic classification is summarized.Second,the classification granularity and its application scenarios are analyzed,and the main technologies of traffic classification are classified according to supervised learning,semi supervised learning and unsupervised learning.Finally,the development trend and challenges of network traffic classification technology are prospected,which provides a certain reference for network traffic analysis and research.

关 键 词：网络流量 机器学习 分类 特征 算法 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]