移动互联网细粒度应用行为识别算法和系统

项目名称： 移动互联网细粒度应用行为识别算法和系统

项目编号： No.61272459

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： 胡成臣

作者单位： 西安交通大学

项目金额： 81万元

中文摘要： 业务识别为移动互联网及其应用的管理、优化和控制提供有效支撑，是下一代互联网发展的必备要求。移动应用的爆炸增长和移动互联网接入速度的显著提高成为精准快速识别移动互联网应用行为的巨大挑战。本课题研究移动互联网迫切需要解决的科学问题，探索细粒度业务行为识别系统的体系结构及其关键算法，采用数学模型和理论求证的手段，通过软件仿真和实验床测试的方法，力图从四方面寻求突破和创新：1）提出高速匹配方法，消除匹配规则间隐藏的冗余度，提高规则集匹配速度；2）深入挖掘规则与规则、流与流之间的依赖关系对规则集进行划分，提高匹配速度；3）提出新的并行识别算法，在保证流亲和性的前提下，利用可乱序负载达到负载均衡，加速整个系统；4）通过对基站与移动设备间的带宽分配原则进行建模，提出小流量永久在线应用的识别方法，为移动基站智能分配带宽提供保证。本课题对提高网络质量和保证网络安全等应用具有重要的理论意义和应用前景。

中文关键词： 网络测量；细粒度业务识别；深度语义检测；；

英文摘要： Application Identification provides efficient support for the management, optimization, and control over mobile Internet and its applications. Explosive growth of mobile applications and the significant improvement of the access bandwidth become the big challenges of accurate and rapid identification of applications and behaviors on mobile Internet. This proposal investigates urgent scientific problem on mobile Internet, seeks the architecture and its key algorithms of a fine-grained traffic behavior identification system, adopts mathematical modeling and theoretical proof, utilizes software simulation and test-bed testing, and try to make breakthrough and innovate in the following four aspects. 1) Design a High-speed matchingmethod to eliminate the redundancy hidden in the rule sets. 2) Mine the dependencies between the rules and the flows, split the rule sets to accelerate the application. 3) Propose a new parallel algorithm, which uses the dis-orderable traffic load to balance the processing, while guaranteeing the affinity of flow. 4) By modeling the interactions between the stations and the mobile devices, propose an algorithm to detect small but long-lived application inorder to guarantee the bandwidth allocation. This proposal has significant theoretical value and is promising to apply to the applications

英文关键词： Network Measurement；Fine-Grained Traffic Identification；Deep Semantics Inspection；；