/

CVE-2026-11718

Severity:

CRITICAL RISK

(9.1)

Authentication Bypass (CWE-287) in Google MCP Toolbox

CVE-2026-11718

Analyzed

Vulnerability type

Category:

Authentication Bypass

CWE:

CWE-287

Impact:

Unauthorized Access

Severity:

CRITICAL RISK

(9.1)

CVSS version:

3.1

Catch this CVE while code is being written

Detect vulnerable code directly in pull requests, so CVEs are fixed before merge.

Summary

A critical authentication bypass in Google MCP Toolbox causes opaque-token validation to skip issuer verification when an OAuth 2.0 introspection response omits the optional iss field.

As a result, the application may accept tokens issued by unauthorized or unintended identity providers, potentially granting access without proper issuer validation.

Affected context

The vulnerability occurs in deployments that:

  • Use the generic opaque-token validation path
  • Validate tokens through an OAuth 2.0 introspection endpoint
  • Configure an expected token issuer
  • Receive introspection responses that omit the optional iss claim

Risk depends on whether an attacker can present a token accepted by the configured introspection service but issued by an unintended identity provider.

Exploit preconditions

  • Attack surface: Authenticated application or API endpoints protected by opaque-token validation
  • Required access: Ability to present a token to the protected service; control of an accepted or reachable identity-provider flow may also be relevant
  • Trigger point: Validation of an opaque token after the introspection response is decoded
  • Why exploitation is possible: The issuer comparison is skipped when the introspection response does not contain an iss value, allowing token acceptance without the intended issuer check

Why this happens

The root cause is an incomplete conditional used during security-claim validation. The implementation checks the configured issuer and the returned issuer only when both values are non-empty, treating a missing issuer claim as a reason to skip validation rather than as a failed or insufficiently trusted response.

This behavior maps to CWE-287: Improper Authentication, because an authentication decision can succeed without verifying a required identity-provider property. The insecure pattern is conditional enforcement of an authentication claim based on attacker-influenced or externally supplied field presence.

Potential impact

  • Unauthorized users may gain access to protected MCP Toolbox functionality
  • Tokens from unintended identity providers may be accepted
  • Authorization boundaries between identity providers could be bypassed
  • Protected data or connected services could be accessed
  • Compromised accounts or tokens could enable further application abuse

General mitigation guidance

Mitigation generally focuses on making issuer validation mandatory whenever an issuer is configured.

Common defensive patterns include:

  • Fail closed when a required introspection claim is missing
  • Require an exact, trusted issuer match rather than conditionally skipping the comparison
  • Treat incomplete or malformed introspection responses as invalid
  • Minimize trust in externally supplied authentication metadata
  • Apply least privilege to identities and downstream MCP tools
  • Separate authentication success from authorization decisions and enforce both explicitly
  • Use defensive tests covering omitted, empty, unexpected, and conflicting issuer values

How teams detect issues like this

Organizations can detect this class of issue through:

  • Static analysis for authentication checks guarded by optional-field presence
  • Code review searches for conditions that skip issuer, audience, or subject validation when values are empty
  • Tests that provide introspection responses with missing or malformed security claims
  • Policy enforcement requiring fail-closed handling for incomplete identity-provider responses
  • Authentication telemetry that records issuer-validation outcomes and unexpected token-provider values

Automated security review tools may help identify conditional claim-validation patterns, but findings should be confirmed against the intended authentication policy.

Sources

SourceLink
github.comgithub.com/googleapis/mcp-toolbox/pull/3360

Frequently asked

questions

Yes, where protected endpoints are exposed and an attacker can submit an opaque token that reaches the vulnerable validation path. Exploitability still depends on the deployment's identity-provider and introspection configuration.

The attacker generally needs a token that the introspection service considers active or otherwise acceptable. The flaw allows that token to bypass the intended issuer restriction; it does not necessarily create a token from nothing.

Require a non-empty issuer claim and compare it against the configured trusted issuer. If the claim is absent, unexpected, or malformed, reject the token rather than skipping validation.

Define required authentication claims explicitly, use fail-closed validation logic, test missing-claim scenarios, and monitor for tokens associated with unexpected identity providers.

Find out what's already
exploitable in your codebase.