There is a growing evidence that aggregated traffic in a variety of network, with different extent and topology, as well as traffic related to specific sources, such as Internet traffic, shows self- similar behaviour. The aim of this paper is to study stochastic properties of a real network traffic trace. From a theoretical point of view, slowly decaying variances, long-range dependence, and spectral density increasing hyperbolically at low frequencies are different manifestations of the self-similar nature of the traffic process. In the analysis of the traffic trace to detect selfsimilar properties we estimate the spectral density function. The behaviour of the spectral density function at low frequencies shows that real network trace exhibits self-similar properties. Then a model of the traffic network trace is found.