CESNET Models

This is documentation of the CESNET Models project.

The goal of this project is to provide neural network architectures for traffic classification and their pre-trained weights. The weights were trained using public datasets available in the CESNET DataZoo package.

The newest network architecture is called 30pktTCNET. It processes packet sequences in order to create flow embeddings that are useful across traffic classification tasks. See the getting started page and models reference for more information.

30pktTCNET

An older network architecture, which apart from packet sequences also utilizes flow statistics, is named Multi-modal CESNET v2 (mm-CESNET-v2).

Multi-modal CESNET v2

Papers

Models from the following papers are included:

• Fine-grained TLS services classification with reject option
  Jan Luxemburk and Tomáš Čejka
  Computer Networks, 2023

• Encrypted traffic classification: the QUIC case
  Jan Luxemburk and Karel Hynek
  2023 7th Network Traffic Measurement and Analysis Conference (TMA)