REAL-TIME IDENTIFICATION OF VoIP APPLICATIONS TRAFFIC

This work addresses the problem of real-time recognition of two VoIP applications (Skype and Zoom) via the analysis of the generated traffic during a call. This type of problem, in fact, lately attracted the attentions of the experts in field since its resolution could prove to be quite useful for a series of purposes, like automatic blocking or filtering of traffic flows or labeling of flows for QoS enforcement. The thesis is structured as follows: after the first chapter that will serve as an introduction, the second chapter will talk about related work, which solutions have already been tested and which levels of accuracy have been reached, and what is the current state of the art; the third chapter will go into more details about the characteristics of the traffic generated by the two applications under analysis; lastly, the fourth chapter will talk about the classifier, how does it work and what are the modules that realize it, while the fifth chapter shows testing results and accuracy values. A conclusive chapter about open issues, possible enhancements and final considerations follows. An appendix was included to briefly explain machine learning fundamentals and the two models used in this work (Decision Tree and Naïve Bayes).