| Version line | Supported |
|---|---|
| 0.5.x | ✅ |
| < 0.5 | ❌ (please upgrade) |
Please do not open a public GitHub issue for security reports. Instead:
- Open a private GitHub Security Advisory at https://github.com/ArioMoniri/syntha/security/advisories — this is the preferred channel.
- Or open a public issue tagged
securityonly if the vulnerability is already public.
We will acknowledge within 5 working days and aim to ship a fix within 30 days of acknowledgement, faster for high-severity issues.
In scope:
- Code execution / data-exfiltration vulnerabilities in the
syntha-ehrpackage or thesyntha-mcpconnector. - Privacy regressions — anything that allows reconstructing a row of the source EHR from the bundled model summaries (the Stadler 2022 NN-MIA gate is the empirical defence; an MIA AUROC > 0.60 result on the released model is in scope).
- Supply-chain risks in the published PyPI / Docker / DMG / EXE / AppImage artifacts.
Out of scope:
- Bugs that affect only research-grade output quality (open a normal issue).
- Vulnerabilities in your own deployment of
syntha-mcp --transport httpif you have failed to terminate TLS, restrict access, or otherwise harden the endpoint. The connector ships with no built-in authentication; it is your responsibility to gate the HTTP endpoint. - Anything in third-party dependencies; please report those upstream and CC us.
Every tagged release runs the Stadler 2022 nearest-synthetic-neighbor MIA
in CI (.github/workflows/privacy-audit.yml); the build fails at AUROC
0.60. If a release ships with a passing CI gate but you can demonstrate a higher AUROC against the bundled model, that is a privacy regression and should be reported via the channels above.
We follow a 90-day coordinated-disclosure window from the date of acknowledgement, extensible by mutual agreement.