Skip to main content

Library for simulating data poisoning attacks against online learning.

Project description PyPI PyPI - Python Version PyPI - License

niteshade (/ˈnaɪt.ʃeɪd/) is a Python library for simulating data poisoning attack and defence strategies against online machine learning systems. The library is written in Python 3.10 and offers a simple and intuitive API which is heavily integrated with PyTorch’s machine learning framework.

For further details about the project, including background information, example usage and detailed API documentation, visit


niteshade requires Python 3.8 or higher.

niteshade binaries may be installed from PyPI using pip

$ pip install niteshade

Use of a virtual environment is strongly recommended.

Note: Installation with Anaconda is not currently available. Anaconda users should install niteshade with pip inside an Anaconda environment.


niteshade is a library of functions and classes which allow users to easily specify data poisoning attack and defence strategies and simulate their effects against online learning using PyTorch models. Detailed information regarding the python API and example usage can be found at


This project has separate package and developer dependencies, both of which can be found in the env/ directory in the form of requirements.txt and environment.yml files for pip and Anaconda users. Package dependencies (prefixed with “pkg”) consist exclusively of the packages required to use the library and are installed automatically when niteshade binaries are installed with pip. Developer dependencies (prefixed with “dev”), include the package dependencies as well as additional packages required for building the documentation, running tests, linting the source code and publishing releases.

$ cd env/

pip users:

$ pip install -r dev_requirements.txt # or pkg_requirements.txt

Anaconda users:

$ conda env create -f dev_environment.yml # or pkg_environment.yml

Building the Documentation

To build documentation in various formats, you will need Sphinx and the readthedocs theme (included in the developer dependencies). You can build the documentation by running make <format> from the docs/ directory. Run make to get a list of all available output formats.

$ cd docs/
$ make clean && make <format>

Running Unit and Integration Tests

This project uses pytest for unit and integration testing (included in the developer dependencies). The tests may be run from the root directory as follows:

$ pytest
===== x passed, x warnings in x.xx seconds =====

Package Releases

Package releases are published to PyPI automatically when a tag is pushed to GitHub.

$ export RELEASE=x.x.x
$ git commit --allow-empty -m "Release $RELEASE"
$ git tag -a $RELEASE -m "Version $RELEASE"
$ git push origin --tags


niteshade was co-created by Mart Bakler, Oskar Fernlund, Alexandra Ntemourtsidou, Jaime Sabal-Bermudez and Mustafa Saleem in 2022 at Imperial College London. The authors may be contacted at the following email addresses:

Many thanks to Dr. Emil C. Lupu for all his insightful feedback and support.

Disclaimer: While niteshade is an open-source project and contributions are welcome, we cannot guarantee that the codebase will be actively maintained in the future.


niteshade uses the MIT license. Details regarding permissions to use and distribute the software may be found in the LICENSE file.

Project details

Download files

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

Source Distribution

niteshade-0.1.4.tar.gz (56.0 kB view hashes)

Uploaded Source

Built Distribution

niteshade-0.1.4-py3-none-any.whl (40.3 kB view hashes)

Uploaded Python 3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page