Report security issues via GitHub. Do not open a public issue.
Expect a reply within a few days, usually within hours. Include:
- The security impact
- A working re-production.
PoCcode is preferred, but a detailed description of the steps to reproduce is acceptable. - The YTPTube release version shown in the UI footer
- Relevant configuration with secrets removed
Security issues in yt-dlp itself, including vulnerabilities in its extractors, option parsing, or command execution, should be reported to the yt-dlp project, not to YTPTube.
Report issues to YTPTube when the vulnerability is caused by its integration or exposes functionality without the required authentication boundary.
Only the latest release is supported. Update to the current Docker image or latest native release before reporting. Reports that cannot be reproduced on the latest release will be closed.
YTPTube is an administrative application, not a sandbox for untrusted users. Anyone with valid application credentials is trusted to manage the entire YTPTube instance and what it has access to.
Reports that only describe the following behavior will be closed:
- Unauthenticated access when
YTP_DISABLE_AUTH=true, including the default configuration in native builds. This setting disables application authentication and is intended for deployments protected by a trusted reverse proxy or private network. - First-user registration through
/setupwhile no account exists. This is the intended bootstrap flow; reports must demonstrate access after setup is complete or another authentication boundary bypass. - Behavior caused by yt-dlp itself, including command execution or path control through options such as
--exec,--output, or--netrc. yt-dlp is not sandboxed and can execute arbitrary commands. - YTPTube does not sandbox outbound network access. Authenticated administrators can cause requests to internal services, including through redirects or yt-dlp behavior or even via dns rebinding.
A report about these areas must demonstrate a boundary bypass, such as access without valid authentication while auth is enabled.
Verify scanner and LLM output before submitting it. Reports that contain unreviewed generated output without a working reproduction and demonstrated impact will be closed.