Entropy-based traffic filtering to support real-time Skype detection
Peter Dorfinger, Georg Panholzer, Brian Trammell, Teresa Pepe · 2010
We propose a novel approach for real-time privacy preserving traffic filtering based on entropy estimation. The decision of the real-time classifier is based on the entropy of the payload from first packet of a flow. The aim of the classifier is to detect traffic with encrypted payload. As a proof of concept we show the applicability of our approach as a traffic filter for a Skype detection engine. Traces collected in laboratory and real-world environments show that the traffic is reduced by a reasonable amount while achieving similar or even improved detection quality.