Amulet
Amulet is a Python machine learning (ML) package to evaluate the susceptibility of different risks to security, privacy, and fairness. Amulet is applicable to evaluate how algorithms designed to reduce one risk may impact another unrelated risk and compare different attacks/defenses for a given risk.
Amulet builds upon prior work titled “SoK: Unintended Interactions among Machine Learning Defenses and Risks” which appears in IEEE Symposium on Security and Privacy 2024. The SoK covers only two interactions and identifies the design of a software library to evaluate unintended interactions as future work. Amulet addresses this gap by including eight different risks each covering their own attacks, defenses and metrics.
Amulet is:
- Comprehensive: Covers the most representative attacks/defenses/metrics for different risks.
- Extensible: Easy to include additional risks, attacks, defenses, or metrics.
- Consistent: Allows using different attacks/defenses/metrics with a consistent, easy-to-use API.
- Applicable: Allows evaluating unintended interactions among defenses and attacks.
Built to work with PyTorch, you can incorporate Amulet into your current ML pipeline to test how your model interacts with these state-of-the-art defenses and risks. Alternatively, you can use the example pipelines to bootstrap your pipeline.
Getting Started
The easiest way to start is via pip:
pip install amuletml
For a reproducible GPU build (and for development), install from source with uv and select the torch build matching your driver: uv sync --extra cu128 (CUDA 12.x), uv sync --extra cu130 (CUDA 13), or uv sync --extra cpu. Add --extra llm for the optional text/LLM stack (used by the textual backdoor pipeline).
Test installation
To test your installation, please run examples/get_started.py. This script also serves as a starting point to learn how to use the library.
Learn More
For more information on the basics about the library, please see the Getting Started guide.
To see the attacks, defenses, and risks (modules) that Amulet implements, please refer to the Module Hierarchy.
For each module, please see amulet/examples for implementations of pipelines that include recommendations on how to run each module.
Contributing
See CONTRIBUTING for guidance.
1 maintainer
Metadata
1 maintainer
Release files for amuletml 0.6.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| amuletml-0.6.0.tar.gz | 13.3 MB | Without provenance | Details for amuletml-0.6.0.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| amuletml-0.6.0-py3-none-any.whl | Python 3 | none | any | Details for amuletml-0.6.0-py3-none-any.whl |
Total release size: 13.4 MB
Release files / amuletml-0.6.0.tar.gz
| Download URL | amuletml-0.6.0.tar.gz |
|---|---|
| Size | 13.3 MB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
eb90cebaa5ae0cdbb8a8d4253b1d65ab4e8556666538eb150620be0f47a0ec4d
|
|
BLAKE2b-256 checksum How to use checksums |
b4de8f9f62d68945cb1823bd6292657d297af83315ec2b8e5092ad8ee83e555e
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
uv/0.11.32 {"installer":{"name":"uv","version":"0.11.32","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"24.04","id":"noble","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}
|
Release files / amuletml-0.6.0-py3-none-any.whl
| Download URL | amuletml-0.6.0-py3-none-any.whl |
|---|---|
| Size | 108.0 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
7d61413ca6eecbfd567e0a3af28aa4f7b7eda2d537cf8e17d8988d197b7bf564
|
|
BLAKE2b-256 checksum How to use checksums |
ad60d0bf789e0656c6cab959e1f0bf4c441c2e6cef438bcdb9696a4bbd757531
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
uv/0.11.32 {"installer":{"name":"uv","version":"0.11.32","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"24.04","id":"noble","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}
|
Security for amuletml 0.6.0
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.
