Noema Atlas handles large model files, manifests, peer discovery, local caches, and user credentials for private sources. Please report suspected vulnerabilities privately so maintainers can investigate before details are public.
Until the project reaches a stable 1.0 release, security fixes target main and
the latest published release. Maintainers may backport critical fixes when a
release is still widely used and a backport is practical.
| Version | Supported |
|---|---|
main |
Yes |
| Latest release | Yes |
| Older pre-1.0 releases | Best effort |
Use GitHub's confidential security advisory form:
https://github.com/noemaai-labs/noema-atlas/security/advisories/new
If private vulnerability reporting is not available, send the report to
clientcare@noemaai.com with Noema Atlas security in the subject.
Please do not open a public issue, pull request, or discussion for suspected vulnerabilities before maintainers have had time to assess the report.
Include as much of the following as possible:
- A clear summary of the issue and affected component.
- Steps to reproduce with synthetic data where possible.
- The affected version or commit.
- Platform details.
- Expected and actual security impact.
- Logs, manifests, or proof-of-concept code with secrets removed.
- Whether the issue is already public or known to be exploited.
Do not attach proprietary model weights, private access tokens, signing keys, or real user data. If the issue requires a large test artifact, describe how to construct a minimal synthetic one.
High-value areas include:
- Manifest signing, canonicalization, and verification.
- Hash verification, quarantine behavior, and source banning.
- P2P, LAN, and Iroh transport behavior.
- Tracker publication, peer discovery, and unintended data exposure.
- Local cache, install, delete, and path handling.
- Secret storage and private source credentials.
- Desktop command surfaces and webview boundaries.
- Release signing, update, and packaging integrity.
- Dependency or supply-chain vulnerabilities that affect shipped code.
Reports are usually out of scope when they require only social engineering, rely on a compromised local administrator account, or describe denial of service from intentionally downloading very large public files without a distinct security boundary failure. Maintainers may still treat these as bugs.
Maintainers aim to acknowledge new reports within two business days, then follow up with triage status, affected versions, and a remediation plan. Timing depends on severity and reproducibility. Coordinated disclosure is preferred; public advisories will credit reporters who want credit.
Good-faith research is welcome when it stays within the scope above, avoids privacy violations, avoids destructive testing, and gives maintainers reasonable time to fix the issue before public disclosure.