中文摘要：

网络流量是局域网和广域网的重要特征之一，小波分析能将复杂的非线性网络流量时间序列分解成不同频率的子序列。基于小波分解的思想，利用网络流量的自相似特性来对网络的异常行为进行检测，给出了根据网络流量自相似特征参数的偏差来检测攻击的方法，对不同分辨率下Hurst参数的变化进行了比较分析。在DARPA上的测试结果表明，该方法不仅能够发现网络中存在的突发性流量攻击，还能够确定异常发生的位置。

英文摘要：

Network traffic is one of the key properties in LAN as well as WAN, by wavelet analysis, the complex traffic times series can be decomposed into different frequent components. Based on wavelet decomposed, the self- similar of traffic can be used to detect anomaly behavior of network, a method of detecting attacks based on the deviation of Hurst parameter is presented. The changes of Hurst parameter are analyzed and compared in different time scale. Result shows the proposed approach can detect the possible presence of not only an anomaly, but also its location on data set.