An attacker with a standard TinaCloud account can exploit this by creating their own application, obtaining a valid token, and submitting requests to a victim site with their own clientID and token. The victim's application incorrectly validates the credentials against the attacker's own app, returning an authorized response. This vulnerability exposes critical functionality, including media bucket management and full GraphQL content read/write/delete capabilities when using the default TinaCloudBackendAuthProvider. The flaw exists across multiple integration libraries, including next-tinacms-cloudinary, next-tinacms-azure, and next-tinacms-dos.

Attack Chain

1. Attacker registers a free TinaCloud account and creates a personal application to obtain a valid clientID and bearer token.
2. Attacker identifies a target self-hosted TinaCMS site that utilizes the vulnerable @tinacms/auth package or its derived media-store integrations.
3. Attacker crafts a malicious HTTP request (e.g., GET /api/cloudinary/media) targeting the victim's API endpoint.
4. Attacker includes their own clientID as a query parameter and their own valid TinaCloud bearer token in the Authorization header.
5. The victim's backend calls isAuthorized(req), which incorrectly performs a look-up at identity.tinajs.io using the attacker-supplied clientID.
6. The identity provider returns a successful validation status because the credentials are valid for the attacker's own app.
7. The victim's backend logic, failing to pin the clientID to the site-specific ID, returns an authorized: true response to the media or GraphQL handler.
8. The attacker performs unauthorized actions, such as reading private media, uploading arbitrary files to the victim's CDN, or deleting/modifying site content.

Impact

Successful exploitation allows an unauthenticated attacker to gain editor-level control over unrelated TinaCMS tenants. Impacts include the ability to list, read, or delete sensitive files within the victim's media bucket. Furthermore, when TinaCloudBackendAuthProvider is active, the attacker gains full GraphQL access, enabling the exfiltration of site content or the injection of malicious data into the CMS.

Recommendation

1. Immediately audit all instances of @tinacms/auth and associated media/backend providers in self-hosted TinaCMS deployments.
2. Implement strict server-side validation to ensure the clientID provided in incoming requests matches the hard-coded or environment-configured application ID for that specific site.
3. Avoid relying solely on the return value of isAuthorized(req) without verifying that the returned user object's associated appId matches the expected deployment identifier.
4. Block or monitor suspicious inbound requests to /api/cloudinary/media or /api/tina/gql where the clientID parameter does not match your organization's known TinaCloud clientID.