Feature Extraction & Modeling Methodology for Next Generation Network Traffic

Traffic classification is a fundamental problem in achieving differentiated services for Next Generation Internet. In this work we propose a statistically based scheme to classify Internet traffic flows over a time frame, based on the application to which the packets belong. The scheme uses statistical information and thus it does not involve reading any packet headers to determine the application. This ‘implicit’ classification builds a foundation for estimation and prediction of traffic mix, which is a long-term goal of this research project. The classification results mentioned in this paper indicate that this approach is promising.