一种改进的基于MMTD的网络应用类型识别方法

Improvement of Network Application Type Identification Based on Measure of Medium Truth Degree

网络管理在很多方面需要识别网络流的应用类型。文中旨在改进和完善一种融合了中介数学理论的基于流统计特征的识别方法———基于MMTD(中介真值程度度量)的应用识别方法。原方法在效率上表现出了优越性和发展潜力,但方法中用到的参数基本都是基于网络专业知识和对流属性值的人工统计分析凭经验确定的。参数的确定没有系统的方法,影响识别方法的使用与推广。文中对原方法进行了如下改进与完善:(1)提出了一种能够反映对象测度(属性)分布特性的测度值与谓词对应关系,并给出了相应的真值程度计算方法;(2)提出了一种基于测度值分布密集程度的测度选取和权重计算方法;(3)提出了一种通过严格的统计及数据处理方法得到各种网络应用流属性的特征参数的方法。实验测试结果表明文中提出的改进和完善方案是有效的。

Internet application identification is important in network management.This paper aims to improve application type identification of Internet flows related to medium mathematics.It proposes a novel application type identification method based on MMTD（Measure of Medium Truth Degree）.The efficiency of the original method shows the potential and advantages.But the parameters used in the method were probably based on network expertise and the experience of the artificial statistics and analysis of features of flows.The lack of systemic method of calculations of parameters affected the use and promotion of the identification method.This paper improves the original method in followings.First,it proposes mappings between truth scales and metrics values that reflect the distribution characteristics of metrics of objects and the corresponding method of calculation of truth scales.Second,it proposes a method of metrics selection and a method of weights calculation based on intensity degrees of distribution characteristics of metrics.Third,it proposes a method that can calculate feature parameters of internet flow attributes by rigorous statistics and data processing methods.The experimental result demonstrates efficiency of the proposed method.