基于决策树的网络流量分类系统的设计与实现
发布时间：2019-06-04 03:35
【摘要】：随着网络用户数量和网络应用规模的快速增长,基于TCP/IP协议的互联网的应用种类越来越多。面对各种具有充分反监测能力的互联网应用的出现,传统的基于端口和应用层载荷特征的识别方法已经难以胜任当前或者将来流量识别的需求。高效、准确、智能、实时地进行互联网流量识别成了一个具有高度挑战性的问题。本文研究了基于决策树的网络流量分类系统的设计与实现,主要工作如下： 首先介绍了基于端口和基于数据深度包的网络流量分类模型的特点及问题,分析了基于机器学习的网络流量分类方法,比较了贝叶斯分类模型、决策树分类模型及其基本思想,并介绍了特性选择采用的常见算法。 其次给出了流量分类系统的流程设计与功能结构,系统包括四个模块,流量采集模块,流量特性统计模块,流量分类模块和分类结果显示模块。之后分析了决策树的构造思路,设计了流量分类系统中的C5.0决策树分类器。 最后给出了流量采集的实现方法、特性统计的实现方法以及实际网络流量分类结果的显示方法,并展示分析了系统的用户界面及对实际网络流量进行分类的结果,验证了系统的有效性。 论文结合机器学习在流量识别研究中的应用,把C5.0决策树应用到网络流量的识别分类中,设计并实现了一个基于决策树的流量分类系统,可实现对实际网络流量进行分类。
[Abstract]:With the rapid growth of the number of network users and the scale of network applications, there are more and more kinds of Internet applications based on TCP/IP protocol. In the face of the emergence of various Internet applications with sufficient anti-monitoring ability, the traditional identification methods based on port and application layer load characteristics have been unable to meet the needs of current or future traffic identification. Efficient, accurate, intelligent and real-time Internet traffic identification has become a highly challenging problem. In this paper, the design and implementation of network traffic classification system based on decision tree are studied. The main work is as follows: firstly, the characteristics and problems of network traffic classification model based on port and data depth packet are introduced. This paper analyzes the network traffic classification method based on machine learning, compares the Bayesian classification model, decision tree classification model and their basic ideas, and introduces the common algorithms used in characteristic selection. Secondly, the flow design and functional structure of the traffic classification system are given. The system includes four modules: traffic collection module, traffic characteristic statistics module, traffic classification module and classification result display module. Then the construction idea of decision tree is analyzed, and the c5.0 decision tree classifier in traffic classification system is designed. Finally, the realization method of traffic collection, the realization method of characteristic statistics and the display method of the actual network traffic classification results are given, and the user interface of the system and the results of classifying the actual network traffic are displayed and analyzed. The effectiveness of the system is verified. Combined with the application of machine learning in traffic identification research, this paper applies c5.0 decision tree to the identification and classification of network traffic, and designs and implements a traffic classification system based on decision tree, which can classify the actual network traffic.
【学位授予单位】：中国科学院大学（工程管理与信息技术学院）
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.06
本文编号：2492445
[Abstract]:With the rapid growth of the number of network users and the scale of network applications, there are more and more kinds of Internet applications based on TCP/IP protocol. In the face of the emergence of various Internet applications with sufficient anti-monitoring ability, the traditional identification methods based on port and application layer load characteristics have been unable to meet the needs of current or future traffic identification. Efficient, accurate, intelligent and real-time Internet traffic identification has become a highly challenging problem. In this paper, the design and implementation of network traffic classification system based on decision tree are studied. The main work is as follows: firstly, the characteristics and problems of network traffic classification model based on port and data depth packet are introduced. This paper analyzes the network traffic classification method based on machine learning, compares the Bayesian classification model, decision tree classification model and their basic ideas, and introduces the common algorithms used in characteristic selection. Secondly, the flow design and functional structure of the traffic classification system are given. The system includes four modules: traffic collection module, traffic characteristic statistics module, traffic classification module and classification result display module. Then the construction idea of decision tree is analyzed, and the c5.0 decision tree classifier in traffic classification system is designed. Finally, the realization method of traffic collection, the realization method of characteristic statistics and the display method of the actual network traffic classification results are given, and the user interface of the system and the results of classifying the actual network traffic are displayed and analyzed. The effectiveness of the system is verified. Combined with the application of machine learning in traffic identification research, this paper applies c5.0 decision tree to the identification and classification of network traffic, and designs and implements a traffic classification system based on decision tree, which can classify the actual network traffic.
【学位授予单位】：中国科学院大学（工程管理与信息技术学院）
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.06
【参考文献】
相关期刊论文 前3条
1 许晴;李凡长;;上下文决策树学习算法及其在机械波图像中的应用[J];合肥工业大学学报(自然科学版);2013年02期
2 徐鹏;林森;;基于C4.5决策树的流量分类方法[J];软件学报;2009年10期
3 赵国锋;吉朝明;徐川;;Internet流量识别技术研究[J];小型微型计算机系统;2010年08期
，本文编号：2492445
本文链接：https://www.wllwen.com/guanlilunwen/ydhl/2492445.html
