/

CVE-2026-84796

Severity:

HIGH RISK

(8.8)

Site Scope Bypass (CWE-639) in Craft CMS GraphQL Mutations

CVE-2026-84796

Deferred

Vulnerability type

Category:

Authorization Bypass

CWE:

CWE-639

Impact:

Cross-Site Unauthorized Access and Modification

Severity:

HIGH RISK

(8.8)

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

Craft CMS GraphQL entry mutation resolvers may accept a caller-supplied siteId without verifying that it is within the token's authorized site scope. An attacker with a token limited to one site could potentially read, modify, or delete entries belonging to other sites by supplying a different site identifier.

Affected context

The issue occurs in Craft CMS deployments that expose GraphQL entry mutation resolvers and use site-scoped tokens or other site-level authorization controls. It becomes risky when mutation arguments are processed without applying the same site-scope authorization checks used for the authenticated token, allowing a client-provided siteId to influence the target site.

Exploit preconditions

  • Attack surface: GraphQL entry mutation endpoints
  • Required access: A valid token or session with access to at least one site
  • Trigger point: Supplying a site identifier in mutation arguments
  • Why exploitation is possible: The resolver does not consistently validate siteId through the centralized argument preparation and authorization path before acting on the requested entry

Why this happens

The underlying bug is an authorization enforcement gap involving a user-controlled object or tenant identifier. GraphQL mutation resolvers allow siteId to reach entry operations without ensuring that it belongs to the authenticated principal's permitted site scope. This behavior maps to CWE-639: Authorization Bypass Through User-Controlled Key, because an attacker can influence the identifier used to select a protected resource.

Potential impact

  • An attacker may read entries from unauthorized sites where relevant read operations are exposed.
  • An attacker could modify content or metadata belonging to another site.
  • An attacker could delete entries outside the token's intended scope.
  • Cross-site content integrity, confidentiality, and availability may be affected.

General mitigation guidance

Mitigation generally focuses on enforcing authorization at the point where a site and entry are resolved, rather than trusting identifiers supplied by the client.

Common defensive patterns include:

  • Deriving the effective site scope from the authenticated principal and server-side policy.
  • Validating every client-supplied siteId against the token's permitted sites before resource lookup or mutation.
  • Applying centralized argument preparation and authorization consistently across all GraphQL resolvers.
  • Using deny-by-default policies for cross-site access and requiring explicit authorization for administrative operations.
  • Separating tenant or site identifiers from unrestricted client input where possible.
  • Auditing tokens, resolver permissions, and mutation operations under least-privilege principles.

How teams detect issues like this

Organizations can look for this class of issue through:

  • Static analysis rules that flag resource lookups or mutations based directly on client-supplied tenant, site, or object identifiers.
  • Code review checks confirming that every resolver applies centralized authorization before object retrieval and mutation.
  • Integration tests that attempt cross-site reads, updates, and deletes using tokens scoped to a different site.
  • API policy enforcement that rejects site identifiers outside the authenticated subject's allowed scope.
  • Audit-log analysis for site identifiers that differ from the caller's authorized scope or for repeated authorization failures.

Automated code review and API security testing tools may help identify inconsistent authorization paths.

Sources

SourceLink
github.comgithub.com/craftcms/cms/security/advisories/GHSA-3wcr-p33w-528f
vulncheck.comvulncheck.com/advisories/craft-cms-5.0.0-rc1-before-5.10.11-graphql-entry-mutation-site-scope-bypass

Frequently asked

questions

It may be exploitable remotely when the GraphQL API is reachable and accepts authenticated entry mutations. Exploitation still depends on the attacker possessing a valid token with access to at least one site.

Not necessarily. A site-scoped token may be sufficient if the resolver accepts a caller-controlled siteId without enforcing the token's site restrictions.

Enforce server-side site authorization for every entry operation and derive or validate the effective site from the authenticated principal rather than trusting mutation arguments.

Use centralized authorization checks, test every tenant or site boundary with negative cases, apply least privilege to tokens, and monitor for cross-scope resource access.

Find out what's already
exploitable in your codebase.