Welcome to TAPAS’s documentation!
TAPAS is a Python library for evaluating the privacy of synthetic data from an adversarial perspective.
Note
This project is under active development.
Thus, the API of each of the modules could change at any time.
Finally, we welcome contributions to our package in any way.
This library is designed to be an extensible privacy toolbox which should be accessible to:
Developers or users of synthetic data generators who want to test their generator against a range of known attacks and a diversity of threat models.
Privacy researchers who want to test and develop new attacks against synthetic data generators.
TAPAS implements a panoply of diverse attacks in order to extract private
information about real datasets from synthetic datasets. Importantly, should
no attack be found to succeed, it does not mean that the generator is safe, as
more sophisticated/specifically tailored attacks might exist. This package thus
mostly aims at probing implementations for known vulnerabilities.
Using the Package
To meet the two use cases above we provide two different interfaces into the package.
The first is a pure Python interface which can be combined directly with
standard python pipelines, see the /examples folder.
The second is a purely command line interface using the tapas command which
is directly installed when you set up the package. This interface allows you to
interact with our package without having to develop in a python ecosystem.
Warning: this is currently unsupported – use Python instead