Network Traffic Based On Wavelet Transform Analysis And Applications

Internet has developed quickly in recent years, but research about Internet does not develop as quickly as possible. It is a hard work to analyze and recognize the complex system structure and variety of Internet. In order to guarantee and improve the quality of service, to advance the development of Internet and to exploit the next generation network, it is necessary to analyze and understand the characteristics of network traffic as much as possible.Self similarity is one of main characteristics of Internet and Hurst parameter is the only parameter to evaluate the self similarity of Internet traffic. In this paper, we analyzed five methods to calculate the Hurst parameter, such as aggregated variance, variance of residuals, periodogram, rescale adjusted range and wavelet. We used these five methods to calculate the Hurst parameter of sample sequences which have certain self similarity. With experiment and analysis, the conclusion is that we can get correct result with these five methods and performance of each method is different, but we can not get the exact value of Hurst parameter with these methods.Traffic model is one part of traffic analysis. This paper mainly analyzed multifractal wavelet model (MWM). We chose four kinds of wavelets, such as Haar, Daubechies, Coiflets and Symlets, and analyzed the influence which the wavelets had on the MWM. With the analysis and comparison between the real traffic and the synthesized traffic, the conclusion is that the synthesized traffic, which was got from the MWM with Haar wavelet, can correctly reflect the statistical characteristics of the real traffic.Attack detection is one of applications of traffic analysis. Based on the careful study of the characteristics of the denial of service (DoS), in this paper, we present a new method, which is based on wavelet multi-resolution analysis to detect the DoS. By calculating the wavelet coefficients' energy at each scale and comparing the energy difference in two successive equal time segments, the method can detect the occurring of the DoS. The method is verified by simulation on NS-2. On the NS-2 platform, we set up a network model, simulated the UDP flood attack and compared the energy difference in two successive equal time segments. The conclusion is that the method can detect the DoS attack in the network..This paper provides theories and examination evidences for network traffic analysis and application. Our research work has the reality meanings for managing network effectively and supplying high quality of service.