This project is part of my Masters Thesis and contains: a small library for quickly producing release mechanisms and a sample program called privpack-app.
In the library one can find: architectures producing bivariate binary data and multivariate gaussian data, it contains code for defining gan criterion by defining
utility and privacy criterions and it contains metric methods to estimate the privacy-leakage by mutual information given some dataset.
The recommended steps to develop on this project include:
- Creating a virtual environment which runs python3.
- Install the package locally with the development packages.
These steps are represented by the following commands:
python3 -m venv venv
source venv/bin/activate
pip install -e .[dev]NOTE: python3 is currently required for this project.
For testing purposes one should use the pytest library included in the development requirements of this package. Create your test in the corresponding test file and run the $ pytest <path/to/test/file> command in the root directory of this project.
An example program is available called privpack-app. To run this program open your python environment and run $ python privpack-app -h in the root folder of this project. This will show all the options available for running the example app. At this moment
two Privacy-Preserving networks have been implemented in the example program: binary, gaussian1 and gaussian5.