Tumult Core

Tumult Core is a programming framework for implementing differentially private algorithms.

For more information, refer to the software documentation and references therein.

A portion of this software, specifically the file tmlt/core/privacy_framework/discrete_gaussian.py, is derived from a work authored by Thomas Steinke [dgauss@thomas-steinke.net], copyrighted by IBM Corp. 2020, licensed under Apache 2.0, and available here at commit cb190d2a990a78eff6e21159203bc888e095f01b. This file, and only this file, is licensed under the Apache 2.0 license. However, all other files and this computer software as a whole are not licensed under the Apache 2 license.

<placeholder: add notice if required>

Overview

• Tumult Core is a privacy engine that automatically provides a proof of differential privacy for any plan that manipulates sensitive data. This framework provides components to transform and measure many differentially private queries.
• All transformation operators have a vetted sensitivity, while all measurement operators are proven to be differentially private.

See CHANGELOG for version number information and changes from past versions.

Testing

To run the tests, install the required dependencies from the test_requirements.txt

pip install -r test_requirements.txt

Fast Tests:

nosetests test/unit test/system -a '!slow'

Slow Tests:

Slow tests are tests that we run less frequently because they take a long time to run, or the functionality has been tested by other fast tests.

nosetests test/unit test/system -a 'slow'

All tests (including Doctest):

nosetests test --with-doctest

See examples for examples of features of Tumult Core.