Network Traffic Classification by Packet Length Signature Extraction

Encrypted network traffic classification has been a topic of research fir many decades now, where the goal is to provide traffic visibility and enhance network management. Recent research work mainly talks about statistical feature based network traffic classification as traditional methods can no longer be used due to its inaccuracy or overhead. In this regard, we propose a packet length signature extraction based approach to classify different classes of traffic such as Audio streaming, Video streaming, Browsing, Chat, P2P etc. In order to achieve this, we have come up with novel feature set to train a J48 decision tree classifier for identifying the classes of network traffic and we also discuss the interpretability of the model.