Pattern recognition approaches for classifying IP flows

Abstract

The assignment of an IP flow to a class, according to the application that generated it, is at the basis of any modern network management platform. However, classification techniques such as the ones based on the analysis of transport layer or application layer information are rapidly becoming ineffective. Moreover, in several network scenarios it is quite unrealistic to assume that all the classes an IP flow can belong to are a priori known. In these cases, in fact, some network protocols may be known, but novel protocols can appear so giving rise to unknown classes. In this paper we propose to face the problem of classifying IP flows by means of different pattern recognition approaches. They have been explicitly devised in order to effectively address the problem of the unknown classes, too. An experimental evaluation of the various proposal on real traffic traces is also provided, by considering different network scenarios. © 2008 Springer Berlin Heidelberg.