State of the Art in Traffic Classification: A Research Review

The Internet, while emerging as the key component for all sorts of communication, is far from well-understood. The goal of traffic classification is to understand the type of traffic carried on the Internet, which continually evolves in scope and complexity. For security and privacy reasons, many applications have emerged that utilize obfuscation techniques such as random ports, encrypted data transmission, or proprietary communication protocols. Further, applications adapt rapidly in the face of attempts to detect certain types of traffic, creating a challenge for traffic classification schemes. Research papers on Internet traffic classification try to classify whatever traffic samples a researcher can find, with no systematic integration of results. With the exception of machine learning techniques for traffic classification, we know of no complete overview of traffic classification attempts. To fill this gap, we have created a structured taxonomy of traffic classification papers and their datasets. To illustrate its utility, we use the taxonomy to answer the recently most popular question about traffic (“How much is peer-to-peer file sharing?”). Our survey also reveals open issues and challenges in traffic classification.