Unleashing the Tiger (pytorch version)

This repository is a simple pytorch version of the original repository in TF-2.

Please give credit to the original repository github and Paper: arxiv that is recently accepted by CCS '21.

Code:

• FSHA_torch.py: It implements the attack and a single-user version of split learning.
• architectures_torch.py: It contains the main network architectures we used in the paper.
• datasets_torch.py: It contains utility to load and parse datasets.

Proof Of Concepts 🐯:

We report a set of jupyter notebooks that act as brief tutorial for the code and replicate the experiments in the paper. Those are:

• FSHA.ipynb: It implements the standard Feature-space hijacking attack on the MNIST dataset.

• Migration of other funcitons in progress.....

Tensorflow/Pytorch Version Comparison:

• main.py: Same as FSHA.ipynb.
• main_tf.py: The original TF-2 version of FSHA.ipynb.

The pytorch version runs slower and has a slighly slower convergence performance.

Cite the work:

@misc{pasquini2020unleashing,
      title={Unleashing the Tiger: Inference Attacks on Split Learning},
      author={Dario Pasquini and Giuseppe Ateniese and Massimo Bernaschi}, 
      year={2020},
      eprint={2012.02670},
      archivePrefix={arXiv},
      primaryClass={cs.CR}
}