基于动态多维特征的网络行为模型研究

项目名称： 基于动态多维特征的网络行为模型研究

项目编号： No.61272447

项目类型： 面上项目

立项/批准年度： 2013

项目学科： 自动化技术、计算机技术

项目作者： 陈兴蜀

作者单位： 四川大学

项目金额： 80万元

中文摘要： 随着大量新兴互联网应用的发展，私有协议、加密技术被广泛采用，使得各种应用的识别愈加困难，给网络管理和监控带来了极大的挑战。项目旨在提出并建立一个适应于不同网络状态的具有自适应能力的网络行为模型，通过模型描述网络应用、主机所表现的行为特征，从而解决网络应用识别、主机追踪的难题。主要内容包括：(a)根据不同的粒度将网络行为分为微观的会话流属性和宏观的主机属性，根据随时间的变化性将网络行为特征分为动态特征和静态特征；(b)对各种网络行为属性进行定义和提取，研究特征抽象和选择方法，对各个属性进行抽象并统一进行形式化描述，选择区分度较大的属性作为网络行为特征；(c)针对特征随时间不断变化的动态性，对网络行为动态特征的抽象和识别方法展开研究；(d)通过对各种粒度的静态和动态特征进行融合，建立基于动态多维特征的网络行为识别模型；(e)研发原型系统对模型的方法和性能进行验证。

中文关键词： 网络行为分析；动态多维特征；网络应用分类；网络用户识别；大数据分析

英文摘要： With the development of a large number of the emerging Internet applications, private protocol and encryption technology is widely used, making the identification of variety of applications more difficult, and thus brought great challenges to network management and monitoring. The project aims to propose a adaptive network behavior model which can adapt to different network status. The model can describe and identify the various network applications and network users, then solve the problem of application identification and host tracing in network. This project will: (a)Divide network behavoir into microscopic behavior of conversation and macroscopic behavior of host in different particle size, and divide the network behavior characteristics into dynamic features and static features according to whether change over time. (b)Define and extract the properties of network behavior, then study the method of how to abstract and select these characteristics, constructive a mathematical model to abstract the various properties and unified formal description, choose properties which have large discrimination as network behavior characteristics. (c)Study the abstraction and identification method for these time-varying dynamic characteristics of network behavior. (d) Integrate static and dynamic characteristics in differe

英文关键词： network behavior analysis；dynamic multi-dimensional feature；network application classification；network host recognition；big data analysis