Combining Multiple Traffic Classification Techniques within a Single Platform

Abstract. Despite the large quantity of papers published in the past few years on network traffic classification, very few implementations have been made available to the scientific community. Moreover, most approaches proposed so far suffer of problems related to the difficulty of assessing their performance and of comparing them. In this chapter we present a software for traffic classification called TIE. which aims at becoming a common tool for the fair evaluation and comparison of different techniques and at fostering the sharing of common implementations and data. It supports the combination of more classification plugins in order to build a multiple classifier system for traffic classification, and its architecture is designed to allow an online usage.