This package provides an implementation of the SIP-GAN generative method. This is a new model for SIP traffic generation that was published in IEEE ISNCC-2021. For simplicity, we refer to this model as SIP-GAN throughout the rest of this document.
figure: adversarial network traffic generation
Any publication that discloses findings arising from using this source code or the model parameters should cite the SIP-GAN paper.
Main contact: Amar Meddahi (amar.meddahi1@gmail.com)
We propose “SIP-GAN” an extension and adaptation of GANs model for SIP (network protocol used for real-time applications), aiming to process and generate SIP traffic at packet level. The proposed generic model includes an encoder, a generator, and a decoder.
Useful resources are available for potential contributors or those interested in the project:
In this GitHub repository you will find the following files:
- code/SIP_INVITE_20000.txt: The dataset used to train the SIP-GAN model. This dataset corresponds to the captured data from 20,000 SIP INVITE requests sent through the network.
- code/toolbox.py: The library containing all the functions useful for the operation of SIP-GAN (in particular the encoder and decoder).
- code/generator.py: The SIP-GAN generator.
- code/dataset.py: A python script showing an example of how to preprocessed the data and use the SIP-GAN encoder/decoder.
This python package is the one that allowed to validate experimentally the approach proposed in the paper. If you have any question about the source code, please contact me.
If you use the code or data in this package, please cite:
@Article{SIPGAN2021,
author = {Amar Meddahi and Hassen Drira and Ahmed Meddahi},
journal = {IEEE 2021 International Symposium on Networks, Computers and Communications
(ISNCC'21)},
title = {{SIP-GAN:} Generative Adversarial Networks for {SIP} traffic generation},
year = {2021},
}SIP-GAN communicates with and/or references the following separate libraries and packages:
We thank all their contributors and maintainers!