On the codes and data sets used in our traffic classification paper,
"Internet Traffic Classification Demystified: Myths, Caveats, and the Best Practices".

Codes

Payload classification code (also includes a module that extracts various flow features for machine learning experiments) and BLINC code

The codes were originally developed by Thomas Karagiannis at Microsoft,
then later modified by Dhiman Barman at Juniper and Hyun-chul Kim at Seoul National University.
Please contact Hyun-chul Kim to request the same payload classification code and/or BLINC code used in our CoNEXT2008 paper,
"Internet Traffic Classification Demystified: Myths, Caveats, and the Best Practices".

PAIX (Palo Alto Internet eXchange-I and II, in DAG format) data set

This data set is not being released or disclosed publicly. This is due to the ambiguity in the law(s) about the privacy of network traffic measurement data. Efforts are underway to clarify these issues and enable a more positive risk assessment in order to balance the community's research needs with legitimate legal risks. This data set is owned and controlled by CAIDA. As a result of the above-concern, access to this data for analysis purposes was done in-person on CAIDA premises.

WIDE data set (WIDE, Keio-I, and Keio-II, in PCAP format)

To access the original WIDE data sets, please contact Kenjiro Cho at IIJ (kjc _at_ iijlab dot net).
It is also possible for us to help you in an indirect way; by letting us get your implementation code, applying it to the data set, and sharing the results.
For more information on this data set, please contact Hyun-chul Kim.

KAIST data set (KAIST-I and KAIST-II, in PCAP format)

Unfortunately, access to this data set is currently not provided. We will update this page when this data set is accessible (only in an indirect way, due to the korean text of law).
For more information on this data set, please contact Sue Moon at KAIST (sbmoon _at_ kaist dot edu).

Last updated: 2009.2.23
Hyun-chul Kim
Email: hyunchulk _at_ gmail dot com