摘 要 近年来互联网与带宽技术的快速发展，使得互联网应用类型变得多样化、私有应用协议逐渐增多，同时互联网应用与智能终端的加速融合以及P2P网络构架的逐步扩大，使得网络用户持续增加、网络流量爆炸性的增长，带宽资源变得日益紧张，极大地影响了用户的上网体验。网络运营商需要对网络流量进行有效的监控去挖掘用户上网行为习惯，从而丰富增值业务类型、提高用户体验，另一方面，网络信息安全风险也日趋严峻，网络流量的合理监控能补充应对各类异常流量情况，从而排除网络安全隐患。因此在复杂的网络环境下，针对这些市场需求和挑战，如何满足人们对差异化服务的需求、有效保证网络环境安全，成为对网络流量可控可管所亟待解决的问题。 本文对现有的流量识别与控制技术进行了分析研究，针对现有流量监控系统识别率与准确率日益低下的问题，设计了一款基于深度包检测(DPI)的改进型的流量监控系统，与传统的流量监控系统相比，识别速度与准确率都得到了一定的提升。 本文主要工作归纳如下： 1.设计了流量监控系统的总体框架图，并阐述了各模块的主要功能以及与其他模块间的交互情况。 2.针对网络应用协议类型以及数据流量中应用层特征分布情况，设计了一种优化的基于协议分组的多种识别机制融合的识别引擎。 3.详细举例描述了特征库文件中不同类型的规则，并设计了不同类型的流量识别流程图。 4.部署了实验室网络，测试了系统的流量识别和控制效果，并对测试结果进行了分析总结。

ABSTRACT In recent years, with the rapid development of Internet and bandwidth technology, Internet applications grow more diversified and private application protocols begin to increase gradually. In the meanwhile, the accelerated converge of Internet applications and intelligent terminals, along with the gradually expanded of P2P network framework, makes the Internet users continue to increase, the network traffic grow explosively and the bandwidth resources become increasingly tense, which greatly affects users’ Internet experience. Thus, network operators need to effectively monitor the network traffic to figure out users’ online behavior habits so as to enrich the value-added business types and improve users’ experience. Furthermore, the risk of network information security is becoming increasingly severe, and the reasonable monitoring of network traffic can help deal with all kinds of abnormal traffic situations, eliminating the hidden dangers of network security. Therefore, it is an urgent issue to how to meet the different service demand proposed by the people and effectively guarantee the security of network environment in terms of traffic management in the complex network environment. In this paper, the existing traffic identification and control technology are analyzed and studied. Responding to the increasingly lower of recognition rate and accuracy rate of current traffic monitoring system, an improved traffic monitoring system based on Deep Packet Inspection(DPI) is designed. Compared with the traditional traffic monitoring system, its recognition speed and accuracy have been improved to a certain extent. The main work of this paper is summarized as follows: 1.The overall framework of traffic monitoring system is designed, and each module’s main function and their interactions with each other are also described. 2.According to the network application protocol types and the application layer feature distribution in data traffic, an optimized recognition engine based on protocol grouping with multiple recognition mechanisms fused is designed. 3.The rules of different types of feature library are described in detail, and the flowchart of different types of flow identification is designed. The lab network is deployed, and the traffic identification control effect of the system is tested. In addition, the test results are also analyzed and summarized.