CVE-2026-84796 in Craft
Summary
by MITRE • 09/02/2026
Craft CMS versions before 5.10.11 contain a site scope bypass vulnerability in GraphQL entry mutation resolvers that fail to validate siteId through ArgumentManager::prepareArguments(). Attackers with tokens scoped to one site can read, modify, or delete entries across unauthorized sites by passing siteId directly in mutation arguments.
If you want to get best quality of vulnerability data, you may have to visit VulDB.
Analysis
by VulDB Data Team • 09/02/2026
Craft CMS versions prior to 5.10.11 are affected by a critical authorization bypass vulnerability within the GraphQL entry mutation resolvers. This flaw stems from an insufficient validation of the site identifier parameter during argument processing. Specifically, the ArgumentManager::prepareArguments() method fails to enforce strict adherence to the scope defined in the user's authentication token when handling mutations that target entries across multiple sites. In a typical multi-site Craft CMS deployment, access control is designed to restrict users or API tokens to specific logical sites within the installation. However, due to this implementation oversight, the system does not cross-reference the siteId provided directly by the client in the mutation arguments against the authorized scope of the active session token.
The technical root cause lies in the order and logic of argument validation. When a GraphQL mutation is executed, such as one intended to update or delete an entry, the resolver accepts input parameters including the target site identifier. Instead of rejecting requests where the requested siteId does not match the scoped permissions associated with the current token, the system proceeds to process the request using the provided value. This allows an attacker who possesses a valid API token restricted to Site A to manipulate data residing in Site B or Site C by explicitly setting the siteId parameter in their GraphQL query payload. The vulnerability effectively neutralizes the multi-tenancy isolation mechanisms built into the CMS, treating all sites as accessible if the correct entry ID is known and the mutation syntax is valid.
The operational impact of this vulnerability is severe for organizations utilizing Craft CMS with multiple distinct websites or environments under a single installation. Attackers can read sensitive content from unauthorized sites, modify critical data such as configuration settings or user profiles, or delete entries entirely. This leads to a complete compromise of confidentiality and integrity across the entire multi-site network. For businesses relying on strict separation between production, staging, and development environments within one CMS instance, this flaw could result in accidental deletion of live content from unauthorized developers or exposure of proprietary information intended for internal use only. The ability to bypass scope restrictions undermines trust in the API layer and exposes the organization to significant data breach risks without requiring elevated privileges beyond a basic scoped token.
To mitigate this risk, administrators must upgrade Craft CMS immediately to version 5.10.11 or later, where the ArgumentManager has been patched to strictly validate that the siteId argument matches the scope of the authenticated user's token. Until an upgrade is performed, organizations should consider restricting GraphQL access at the network level if possible and auditing API usage for anomalous patterns involving cross-site mutations. This vulnerability aligns with CWE-269 Improper Privilege Management as it involves a failure to enforce proper authorization checks on resource access based on user roles or scopes. Furthermore, from an offensive security perspective, this behavior is consistent with ATT&CK technique T1078 Valid Accounts, where attackers leverage legitimate credentials but exploit misconfigurations in permission boundaries to expand their access beyond intended limits. Regular updates and strict adherence to version control policies are essential defenses against such logic-based authorization flaws.