Skip to main content

Robust ML API

Project description

Interfaces for defining Robust ML models and precisely specifying the threat models under which they claim to be secure. Also includes interfaces for specifying attacks and evaluating attacks against models.

The motivation behind this project is to make it easy to make specific, testable claims about the robustness about machine learning models. Read more in the FAQ.

Installation

You can install from PyPI: pip install robustml.

Usage

See this repository for a complete example of implenenting a model, implementing an attack, and evaluating the attack against the model.

If you’re implementing a defense, you should implement robustml.model.Model. See here for an example.

If you’re implementing an attack against a specific defense, you should implement robustml.attack.Attack. See here for an example.

To evaluate a specific attack against a specific defense, use robustml.evaluate.evaluate(). See here for an example.

Contributing

Do you have ideas on how to improve the robustml package? Have a feature request (such as a specification of a new threat model) or bug report? Great! Please open an issue or submit a pull request.

Before contributing a major change, it’s recommended that you open a pull request first and get feedback on the idea before investing time in the implementation.

Packaging

  1. Update version information.
  2. Build the package using python setup.py sdist bdist_wheel.
  3. Sign and upload the package using twine upload -s dist/*.
  4. Create a signed tag in the git repo with the version number that was uploaded to PyPI.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for robustml, version 0.0.3
Filename, size File type Python version Upload date Hashes
Filename, size robustml-0.0.3-py3-none-any.whl (9.0 kB) File type Wheel Python version py3 Upload date Hashes View
Filename, size robustml-0.0.3.tar.gz (6.7 kB) File type Source Python version None Upload date Hashes View

Supported by

Pingdom Pingdom Monitoring Google Google Object Storage and Download Analytics Sentry Sentry Error logging AWS AWS Cloud computing DataDog DataDog Monitoring Fastly Fastly CDN DigiCert DigiCert EV certificate StatusPage StatusPage Status page