Discovering of Traffic Types based on Statistical Properties

Nowadays many kinds of common and new application generate numerous types of data flows making Internet traffic very heterogeneous. All these traffics gets usually the same service quality – e.g. best effort – however among Internet Service Providers there is an upcoming need to differ users’ traffic somehow in order to provide different service quality for different applications, or filter out some undesirable traffic types. Because of traffic encryption and arbitrary usage of port numbers, identification of data flows became a very hard task. This paper presents an idea for traffic identification, with which automatic traffic classification or even priority employment could be carried out. The method uses statistical traffic parameters for identification, those sources from a network link monitoring tool. This probe registers the arriving packets first ‘n’ bytes with timestamps, from which packet- and transport level properties retrievable. The identification is based on comparing the examined traffic properties to the previously recorded samples that will be also presented in the article through some examples.