Start with Identity
Identity CVE · SAML

CVE-2025-66567ruby-saml bypass after 1.12.4, PortSwigger Fragile Lock

critical
Product: ruby-samlVendor: SAML-ToolkitsCWE-347Disclosed: 2025-12-08Status: PatchedProtocol deep diveNVD ↗

What broke

PortSwigger's December 2025 research, "The Fragile Lock," showed that ruby-saml versions below 1.18.0, including 1.12.4 (the first patched line for the March bugs), still accepted forged SAML assertions. CVE-2025-66567 and CVE-2025-66568 are that pair. Fixed in ruby-saml 1.18.1.

Why it matters

A year of "we patched ruby-saml" tickets was not enough. Anyone who stopped at 1.12.4 stayed exposed through the 2025 holiday change freeze, which is exactly when identity incidents are hardest to staff.

What to do

  • Pin ruby-saml >= 1.18.1. Reject 1.12.4 as "patched" in any SBOM or Dependabot baseline.
  • If GitLab, omniauth-saml, or an internal Rails SP was on the 1.12 line in December 2025, assume the ACS was exploitable and review admin SSO logs for unexpected first-time NameIDs.
  • Read the SAML 2.0 deep dive pitfalls section before you write the next custom verifier.

After you patch

A SAML bypass means the service provider accepted an assertion it should have rejected, so anyone who exploited it authenticated as a real user and left a normal-looking log line.

  • Revoke every session issued by the affected service provider, then rotate its session signing keys. Patching stops new forgeries and does nothing about sessions already minted.
  • Audit administrative accounts and group memberships for changes during the exposure window. Signing in as an administrator is the point of this class, and adding a second account is the standard persistence step.
  • Rotate the identity provider signing certificate if the flaw involved signature validation, and confirm the service provider pins the expected certificate rather than trusting anything in the assertion.
  • Check your own implementation for the same class: exact-match comparison on verification results, rejection of unexpected signature algorithms, and audience and recency checks on every assertion. See SAML 2.0 and SAML vs OIDC.

Sources

Last reviewed By SWI Community TeamSuggest a correctionHow we research

Technique

This CVE is an instance of Federation trust abuse and SAML forgery. A service provider that accepts a SAML assertion it should have rejected treats a forged identity as authenticated, because the failure sits in signature validation code, not in cryptography.

Know a primary source we should add, or a patch status that has changed? Email community@startwithidentity.com. See all briefs in the identity CVE catalog, or volunteer as a CVE Analyst.
Compiled from vendor advisories, NVD, CISA KEV, and public research. CVSS figures can disagree across NVD and the CNA. Confirm affected versions against the vendor advisory before you patch. Independent, community-driven analysis. See the disclaimer.