TrafficS: A Behavior-Based Network Traffic Classification Benchmark System with Traffic Sampling Functionality

摘要

In recent years, there have been many methods proposed to perform network traffic classification based on application protocols. Still, there is a pressing need for a practical tool to benchmark the performance of these approaches in real-world high-performance network environments. In this paper, based on rigorous requirements analysis on real-world environments, we present a real-time traffic classification benchmark system, termed Traffics, which aims at easy performance-evaluation between different intelligent methods. Traffics is not only extensible to incorporate multiple traffic classification engines but supports different packet/stream sampling techniques as well. Furthermore, it could provide users a comprehensive means to perceive the difference between inspected methods in various aspects.