Security fixes ship against the latest minor release only. Upgrade to its newest patch when a security release lands.
| Version | Supported |
|---|---|
| Latest minor release | Yes |
| Older releases | No |
Please do not open a public GitHub issue for security reports.
Report privately via GitHub's Security Advisories form. The advisory is private until a fix is published, and lets us coordinate disclosure, request a CVE if appropriate, and credit the reporter.
If you cannot use the GitHub form, email the maintainer instead: fangchen.li@outlook.com (PGP not currently published — open an advisory and request a key if you need one).
- The affected component (notebook UI, core data plane, a specific driver adapter, the release/install path, etc.)
- A minimal reproduction or proof-of-concept
- The threat model you're assuming (personal deployment, service deployment, multi-tenant, etc. — see deployment modes)
- The Strata version (
strata-notebook --versionorpip show strata-notebook) and Python version
- Acknowledgement within 3 business days.
- A triage decision (accepted / declined / needs-more-info) within 10 business days.
- For accepted issues, a target fix window depending on severity:
- Critical (RCE, auth bypass, credential disclosure): patch release within 7 days where feasible.
- High: patch release within 30 days.
- Medium / low: rolled into the next regular release.
We will credit reporters in the release notes unless you ask us not to.
In scope:
- The published
strata-notebookwheel on PyPI and its console scripts (strata-notebook,strata,strata-notebook-tui,strata-worker). - The HTTP and WebSocket APIs documented in the repo.
- The bundled frontend SPA.
Out of scope (please do not report these as vulnerabilities):
- Personal-mode deployments bound to a non-loopback address without
allow_remote_clients_in_personal=True— Strata refuses to start in that configuration and the guard is the security boundary. - Cache content visible across principals — by design (see CLAUDE.md "Cache is shared across principals"); ACL gates request access, not cache contents.
- Issues that require an attacker who already controls the executor process or the notebook's underlying Python environment.
- Findings produced by static-analysis tools without a concrete exploit path (we triage these separately under our internal CodeQL backlog).