zhaoyongrui / mtc Goto Github PK

This is a development project based on an existing work. Following the model architecture, parameters, and utilizing some datasets mentioned in the original paper, the goal is to implement the experimental results. However, performance matching the paper cannot be guaranteed. For simplicity in development, this project exclusively utilizes the ISCX VPN-non VPN dataset and follows the preprocessing methods outlined in the paper. Finally, the model is trained according to the parameters specified in the paper.

Build the Python environment on either cloud or on-premises machines. To facilitate model training, please follow these simple example steps.

• Install python packages

  pip install requirements.txt

• Download ISCX VPN-non VPN dataset from here.

• Preprocess ISCX VPN-non VPN data through a Makefile

  • Several pickle files would be generated in data/

    make preprocess

• To execute various model training tasks through a Makefile, follow the step below:

  • Execute all training jobs for each model

    make train-all

(back to top)

• Multi-task model Implementation
  • 1D-CNN model
  • Transformer model
  • MTC
    • 1D-CNN + transformer + fusion blocks

(back to top)

Distributed under the MIT License. See LICENSE for more information.

• K. Wang, J. Gao and X. Lei, "Mtc: a multi-task model for encrypted network traffic classification based on transformer and 1d-cnn," Intelligent Automation & Soft Computing, vol. 37, no.1, pp. 619–638, 2023.

• W&B Report

• ISCX VPE-non VPN

  • After filtering by application type

    Application	Amount
    aim_chat	2366
    email	19705
    facebook	2472071
    ftps	4378
    gmail	5242
    hangouts	4419276
    icq	2490
    netflix	474
    scp	19270
    sftp	1351
    skype	3727604
    spotify	939
    torrent	6885
    vimeo	611
    voipbuster	1559956
    youtube	2230