/

CVE-2026-90579

Severity:

HIGH RISK

(7.3)

Missing Authentication (CWE-287) in Cheshire Cat AI

CVE-2026-90579

Deferred

Vulnerability type

Category:

Missing Authentication

CWE:

CWE-287, CWE-306

Impact:

Authentication Bypass

Severity:

HIGH RISK

(7.3)

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 high-severity authentication vulnerability in Cheshire Cat AI affects the HTTP key authorization logic and may allow a remote attacker to bypass authentication by manipulating the user identifier input.

The issue becomes risky when the affected authorization handler processes attacker-controlled identity data without reliably verifying that the request is authenticated.

Affected context

The vulnerability occurs in Cheshire Cat AI deployments using the custom authentication handler in core/cat/factory/custom_auth_handler.py. It affects the _authorize_http_key function when remotely supplied user identity information is accepted without sufficient authentication checks. The issue is associated with versions up to 1.9.2, although exposure depends on whether the affected HTTP authentication path is enabled and reachable.

Exploit preconditions

  • Attack surface: Remotely reachable HTTP authentication endpoint
  • Required access: Network access to the affected service; valid credentials may not be required if the authorization check can be bypassed
  • Trigger point: Processing of attacker-controlled user_id data by _authorize_http_key
  • Why exploitation is possible: The authorization flow may accept or derive identity information without reliably proving that the request corresponds to an authenticated principal

Why this happens

The underlying bug is a failure to enforce authentication before trusting identity information supplied through an HTTP request. If a user-controlled identifier is treated as sufficient evidence of identity, an attacker may access protected functionality without completing the intended authentication process.

This behavior maps primarily to CWE-287: Improper Authentication and is also related to CWE-306: Missing Authentication for Critical Function, because a security-sensitive operation may be reachable without an effective authentication gate.

Potential impact

  • An unauthenticated attacker may access protected application functionality
  • Attackers could read, modify, or delete data exposed through authorized endpoints
  • The vulnerability may enable unauthorized administrative or user-level actions
  • Compromised access could support further attacks against connected services
  • Security monitoring and audit attribution may become unreliable

General mitigation guidance

Mitigation generally focuses on making authentication mandatory and independently verifiable for every protected operation.

Common defensive patterns include:

  • Deriving the authenticated identity from a validated credential or trusted session rather than directly from request parameters
  • Rejecting missing, malformed, expired, or unverifiable authentication material by default
  • Applying centralized authorization checks to every security-sensitive route and function
  • Enforcing least privilege for authenticated identities and service accounts
  • Separating authentication from user-supplied identity or profile metadata
  • Avoiding detailed authentication error responses that reveal which identity values are accepted

How teams detect issues like this

Organizations can look for this class of issue through:

  • Static analysis rules identifying authorization decisions based directly on request parameters
  • Code review checks for handlers that accept a user identifier before validating credentials or sessions
  • Tests that exercise protected endpoints with missing, altered, expired, and mismatched authentication data
  • Policy enforcement requiring centralized authentication middleware for sensitive routes
  • Monitoring for successful requests lacking expected authentication evidence or showing inconsistent identity information

Automated security review tools may help identify missing authentication gates, but manual review is useful for confirming trust boundaries and authorization flow.

Sources

SourceLink
github.comgithub.com/cheshire-cat-ai/core/issues/1137
vuldb.comvuldb.com/cve/CVE-2026-90579
vuldb.comvuldb.com/submit/913219
vuldb.comvuldb.com/vuln/403164
vuldb.comvuldb.com/vuln/403164/cti

Frequently asked

questions

Yes, the reported attack surface is remote. Exploitability depends on whether the affected HTTP authentication path is reachable and exposed in the deployment.

The reported weakness may allow authentication bypass, so valid credentials may not be required. Network access to the affected service and a request that reaches the vulnerable authorization logic are still generally necessary.

Require successful validation of a credential, token, or trusted session before selecting an identity or performing a protected operation. Do not treat a client-supplied user identifier as proof of authentication.

Use centralized authentication middleware, enforce default-deny authorization, test unauthenticated and identity-confusion cases, and review every security-sensitive function for an explicit authentication requirement.

Find out what's already
exploitable in your codebase.