NMN follows semantic versioning. Security fixes are applied to the latest minor release.
| Version | Supported |
|---|---|
| 0.3.x | ✅ |
| < 0.3 | ❌ |
Please do not report security vulnerabilities through public GitHub issues.
Instead, report them privately by emailing taha@azetta.ai with the subject line [SECURITY] nmn.
Please include as much of the following as you can:
- A description of the issue and its potential impact
- Steps to reproduce, including code, inputs, and environment (OS, Python version, framework version)
- Whether the issue is already public or known elsewhere
- Whether you would like to be credited in the disclosure
You should receive an acknowledgment within 72 hours. We will follow up with a more detailed response within 7 days indicating next steps.
- We will work with you to understand and reproduce the issue.
- We will prepare a fix in a private branch and coordinate a release.
- Once a fix is released, we will publicly disclose the vulnerability via a GitHub Security Advisory and a release note in
CHANGELOG.md. - We aim for 90 days from report to public disclosure as a maximum; faster if a patch is straightforward.
The following GitHub repository settings are part of NMN's release security boundary and must remain enabled:
- the
masterbranch ruleset requires pull requests and the repository's lint, type-check, base-install, and representative backend test checks; - deletion and non-fast-forward updates are blocked for
master; - version tags matching
v*.*.*cannot be deleted or updated; - secret scanning, secret-scanning push protection, and Dependabot security updates are enabled;
- merged pull-request branches are deleted automatically; and
- the
pypienvironment retains its required reviewer and trusted-publisher configuration.
The publication workflow provides a second release boundary: a version tag is
accepted only when its commit is contained in origin/master, artifacts are
built once, TestPyPI publication succeeds, and the protected pypi environment
approves the final OIDC publication. Repository administrators should audit
these settings after ownership, plan, or workflow changes.
In scope:
- The
nmnPython package and its public API surface - Build, packaging, and release configuration (
pyproject.toml,.github/workflows/) - Documentation that could lead to insecure usage
Out of scope:
- Vulnerabilities in upstream dependencies (PyTorch, JAX, TensorFlow, etc.) — report those to their respective projects
- Issues in user-provided code or model checkpoints
- Theoretical numerical-stability issues that do not lead to a security impact
Thank you for helping keep NMN and its users safe.