Internet网络流量的自相似性分析

Self-similarity Analysis of Internet Network Traffic

网络流量的自相似参数估计方法有多种,但研究表明这些方法在准确性或计算上都有一定的局限性。该文在小波包分解的基础上,考虑信号能量在分解过程中的分布,得到基于小波包分解的Hurst参数估计方法。通过对两组合成数据的参数估计,表明该方法在继承了小波变换的计算优势的基础上,能得到更精确的估计结果。把该方法应用于计算实际互联网流量的自相似参数,并在此基础上着重分析了受到蠕虫攻击的Internet流量的自相似性的变化,得到了一些有用的结论。

There are several methods to estimate parameter H, which measures the degree of self-similarity of network traffic, but researches have indicated that they have limitations either on accuracy or on computation. This paper introduces a method based on the discrete wavelet packet transform （DWPT） for Hurst parameter estimation. It estimates the synthesis data and real lnternet data by this method for the purpose of validating the accuracy and robustness. Then it focuses on the variation of self-similarity of abnormal traffics with different lnternet worms by the parameter H and the percentage of component grouped by protocol types.