Research on characterization of wireless LANs traffic

In this paper, we employ actual wireless data that draw from well known archives of network traffic traces and investigate the characterization of the wireless LANs traffic. Firstly, useful preliminary information regarding the general wireless traffic dynamics is obtained using one standard statistical technique named Fourier power spectrum. Then the estimation of the parameters, such as the correlation dimension, the largest Lyapunov exponent and the principal components analysis indicate the existence of low-dimensional deterministic chaos in wireless traffic time series. Our results also show that the parameters selection of the phase space reconstruction influence the value of the correlation dimension and the largest Lyapunov exponent, but can not influence on diagnosis of chaotic nature of wireless traffic.

Research highlights
► As the popularity of WLAN grows at an unprecedented rate, the performance of the wireless service in WLAN becomes an important issue. The performance of various mechanisms and policies, which have been proposed to achieve good performance, to various extents, depends on the network’s traffic characteristics. This paper describes an in-depth study on the characteristics of WLAN traffic.
► We propose to use the Chaos Theory to more accurately analyze wireless traffic. The main and creative contributions include the following aspects:
► We employ three actual wireless traces and investigate the characterization of the WLAN traffic. One trace was collected from the WLAN at Tianjin University. The other two traces draw from well known archives of network traffic traces.
► The general wireless traffic dynamics is obtained using one standard statistical technique named Fourier power spectrum. The comparison of power spectrum for three wireless traffic, noise, periodic, and chaotic time series (Chua’s Circuit) is discussed. The power spectrum of wireless traffic not only exhibits sharp spectral lines (or peaks) at certain frequencies but also is somewhat continuous (broadband) which are regarded as a possible indicator of chaos in the wireless traffic.
► The correlation dimension, the largest Lyapunov exponent and the principal components analysis (PCA) method are employed for comprehensive characterization of the wireless traffic dynamics. It is found out that the correlation dimension is non-integer number. We used the small data sets algorithm to calculate the largest Lyapunov exponent. It appeared that the largest Lyapunov exponent is positive. The principal components analysis showed that the intrinsic information of the traffic is accumulated in the first and few lower-index components. All those results indicated that the wireless traffic is a low dimensional chaotic system.
► Comparing to the autocorrelation function method and Takens’ embedding theorem, the mutual information method and Cao’s algorithm is employed to select the optimal embedding delay and the embedding dimension respectively. The results show that these parameter selections influence the value of the correlation dimension and the largest Lyapunov exponent, but can not influence on diagnosis of chaotic nature of wireless traffic.