Skip to main content

Weblate EUVDEUVD-2025-203462

| CVE-2025-66407 MEDIUM
Cross-Site Request Forgery (CSRF) (CWE-352)
2026-05-26 https://github.com/WeblateOrg/weblate GHSA-hfpv-mc5v-p9mm PYSEC-2025-231
5.0
CVSS 3.1 · GitHub Advisory
Share

Severity by source

GitHub Advisory PRIMARY
5.0 MEDIUM
AV:N/AC:L/PR:L/UI:N/S:C/C:L/I:N/A:N

Primary rating from GitHub Advisory · only source for this CVE.

CVSS VectorGitHub Advisory

CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:C/C:L/I:N/A:N
Attack Vector
Network
Attack Complexity
Low
Privileges Required
Low
User Interaction
None
Scope
Changed
Confidentiality
Low
Integrity
None
Availability
None

Lifecycle Timeline

2
Source Code Evidence Fetched
Jun 08, 2026 - 11:39 vuln.today
Analysis Generated
Jun 08, 2026 - 11:39 vuln.today

DescriptionGitHub Advisory

Impact

The Create Component functionality in Weblate allows authorized users to add new translation components by specifying both a version control system and a source code repository URL to pull from. However, the repository URL field is not validated or sanitized, allowing an attacker to supply arbitrary protocols, hostnames, and IP addresses, including localhost, internal network addresses, and local filenames.

When the Mercurial version control system is selected, Weblate exposes the full server-side HTTP response for the provided URL. This effectively creates a server-side request forgery (SSRF) primitive that can probe internal services and return their contents. In addition to accessing internal HTTP endpoints, the behavior also enables local file enumeration by attempting file:// requests. While file contents may not always be returned, the application’s error messages clearly differentiate between files that exist and files that do not, revealing information about the server’s filesystem layout.

In cloud environments, this behavior is particularly dangerous, as internal-only endpoints such as cloud metadata services may be accessible, potentially leading to credential disclosure and full environment compromise.

Patches

This has been addressed in the Weblate 5.15 release.

  • https://github.com/WeblateOrg/weblate/pull/17103
  • https://github.com/WeblateOrg/weblate/pull/17102

Workarounds

Removing Mercurial from VCS_BACKENDS avoids this vulnerability, as the Git backend is not affected. The Git backend was already configured to block the file protocol and does not expose the HTTP response content in the error message.

References

Thanks to Jason Marcello for responsible disclosure.

AnalysisAI

Server-side request forgery in Weblate's Create Component feature allows authenticated users with component creation rights to make the Weblate server issue HTTP requests to arbitrary internal endpoints, including cloud metadata services, by supplying a crafted repository URL when the Mercurial VCS backend is selected. The Mercurial backend uniquely surfaces the full HTTP response body to the requesting user, enabling enumeration of internal services; file:// scheme requests additionally allow filesystem layout inference via error-message oracle behavior. No active exploitation has been confirmed (not in CISA KEV), but cloud-hosted deployments face elevated risk from IAM credential disclosure via instance metadata endpoints.

Technical ContextAI

Weblate is an open-source web-based translation management system distributed as a Python pip package (pkg:pip/weblate). The vulnerability resides in the component creation workflow: the repository URL field accepted arbitrary user-supplied input without scheme or hostname validation. When Mercurial (hg) is the selected VCS backend, Weblate's application server makes an outbound connection to that URL and reflects the HTTP response body back to the user - behavior absent from the Git backend, which had preexisting file:// protocol blocking and does not expose response content on error. The metadata lists CWE-352 (Cross-Site Request Forgery), which appears to be a misclassification; the advisory, researcher disclosure, and tags all identify this as CWE-918 Server-Side Request Forgery (SSRF). The fix, delivered in PRs #17102 and #17103, replaced the permissive VCS_FILE_PROTOCOL boolean with two new allowlist settings - VCS_ALLOW_SCHEMES (defaulting to https and ssh only) and VCS_ALLOW_HOSTS - enforced via a new weblate.utils.validators.validate_repo_url Django field validator applied to the repo and push URL model fields.

RemediationAI

Upgrade to Weblate 5.15, the vendor-confirmed fix release, which introduces VCS_ALLOW_SCHEMES (defaulting to https and ssh) and VCS_ALLOW_HOSTS allowlists enforced at the model validation layer. Docker deployments should configure the corresponding WEBLATE_VCS_ALLOW_SCHEMES and WEBLATE_VCS_ALLOW_HOSTS environment variables post-upgrade. Patch pull requests: https://github.com/WeblateOrg/weblate/pull/17102 and https://github.com/WeblateOrg/weblate/pull/17103. If immediate upgrade is not feasible, the vendor-documented workaround is to remove Mercurial from the VCS_BACKENDS configuration setting - this fully eliminates the attack surface because the Git backend is unaffected. Trade-off: removing Mercurial will break any existing translation components currently using the Mercurial backend, requiring migration to Git before applying the workaround. As an additional layer for cloud-hosted instances, enabling IMDSv2 token requirements on AWS EC2 (or equivalent hop-token enforcement on GCP/Azure) will block metadata endpoint access even if SSRF is triggered, reducing the most severe impact scenario.

Share

EUVD-2025-203462 vulnerability details – vuln.today

This site uses cookies essential for authentication and security. No tracking or analytics cookies are used. Privacy Policy