Skip to content

Modeled discriminative prior problem learning privacy-utility trade-off Private Information Bottleneck un-supervised

License

Notifications You must be signed in to change notification settings

BehroozRazeghi/Variational-Leakage

Repository files navigation

GitHub Repo license GitHub Repo stars GitHub issues

Description

The official code of "Variational Leakage: The role of Information Complexity in Privacy Leakage" [link].

Getting Started

Dependencies

  • Python 3.8.X
  • TensorFlow 2.4.1+
  • Pandas
  • Matplotlib (for graphs and figures)
  • OpenCV-python (to read and preprocess image data in CelebA experiments)

Installing

  • Download repository
  • Install Dependencies
  • Download img_align_celeba.zip and extract it to the CelebA folder in the root of the project

Executing program

  • Run data_colored_mnist.ipynb to generate Colored-MNIST dataset (both biased and uniform)

  • Run exp_colored_mnist.ipynb for Colored-MNIST experiments of the paper

  • Run exp_celeba.ipynb for CelebA experiments of the paper

Help

It is highly recommended to select the appropriate batch size based on your GPU's memory and computer specification to utilize maximum efficiency on your computer.

Authors

  • Amir Atashin
  • Behrooz Razeghi

Version History

  • 1.0
    • Initial Release

License

This project is licensed under the MIT License - see the LICENSE file for details

Citation

Please consider referencing the following research paper of this repository if you find it useful or relevant to your research:

@inproceedings{10.1145/3468218.3469040,
author = {Atashin, Amir Ahooye and Razeghi, Behrooz and G\"{u}nd\"{u}z, Deniz and Voloshynovskiy, Slava},
title = {Variational Leakage: The Role of Information Complexity in Privacy Leakage},
year = {2021},
isbn = {9781450385619},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
url = {https://doi.org/10.1145/3468218.3469040},
doi = {10.1145/3468218.3469040},
abstract = {We study the role of information complexity in privacy leakage about an attribute of an adversary's interest, which is not known a priori to the system designer. Considering the supervised representation learning setup and using neural networks to parameterize the variational bounds of information quantities, we study the impact of the following factors on the amount of information leakage: information complexity regularizer weight, latent space dimension, the cardinalities of the known utility and unknown sensitive attribute sets, the correlation between utility and sensitive attributes, and a potential bias in a sensitive attribute of adversary's interest. We conduct extensive experiments on Colored-MNIST and CelebA datasets to evaluate the effect of information complexity on the amount of intrinsic leakage.},
booktitle = {Proceedings of the 3rd ACM Workshop on Wireless Security and Machine Learning},
pages = {91–96},
numpages = {6},
keywords = {Information complexity, statistical inference, intrinsic leakage, privacy, information bottleneck},
location = {Abu Dhabi, United Arab Emirates},
series = {WiseML '21}
}

Acknowledgments

Inspiration, code snippets, etc.