RBI Assistant Platform CVE-2025-62647
MEDIUMSeverity by source
AV:N/AC:L/PR:N/UI:N/S:C/C:N/I:L/A:N
Unauthenticated network API with low complexity (PR:N/AC:L); scope change (S:C) reflects cross-store authorization boundary crossing; impact limited to integrity via file upload, so C:N/I:L/A:N.
Primary rating from NVD.
CVSS VectorNVD
Lifecycle Timeline
2DescriptionNVD
The Restaurant Brands International (RBI) assistant platform through 2025-09-06 provides the functionality of returning a JWT that can be used to call an API to return a signed AWS upload URL, for any store's path.
AnalysisAI
Broken authorization in the Restaurant Brands International (RBI) 'assistant' platform (as it operated through 2025-09-06) lets unauthenticated remote attackers request a JWT and then supply an arbitrary store's path to an API that returns a signed AWS upload URL, enabling file uploads into any store's storage location rather than only their own. Per the assessed vector (AV:N/AC:L/PR:N/UI:N/S:C/C:N/I:L/A:N) the flaw is unauthenticated with no user interaction and is scoped to integrity impact only, with no direct confidentiality or availability loss. Publicly available exploit code exists and researchers have publicly demonstrated the issue, though EPSS is low at 0.38% (29th percentile) and there is no CISA KEV confirmation of active exploitation.
Technical ContextAI
The root cause is CWE-863 (Incorrect Authorization), specifically an insecure direct object reference in which a JWT issued by the RBI assistant platform is accepted by a downstream API that mints a signed (presigned) AWS S3 upload URL. Because the API does not bind the requested store path to the identity or store scope encoded in the authenticated token, a caller can substitute any store's path and receive a URL that authorizes writes into that location. The weakness sits at the intersection of token issuance, authorization policy enforcement, and cloud storage delegation: the JWT proves some level of system access but no per-store authorization check is performed before the presigned URL is generated. CPE data (cpe:2.3:a:rbi:restaurant_brands_international_assistant:*:*:*:*:*:*:*:*) indicates the affected component is RBI's own proprietary platform with no explicit version bound, which is consistent with a single hosted service rather than broadly distributed installable software; the weakness may plausibly have been remediated server-side after the 2025-09-06 observation date without any published version identifier.
RemediationAI
No vendor-released patch with an identifiable version number has been published; the remediation is a server-side authorization fix, and operators of the platform should assume the vulnerable behavior persists until RBI confirms otherwise. The correct fix is to bind every signed-upload-URL request to the store scope asserted by the presented JWT, rejecting any request whose target path falls outside the caller's authorized store identifier, and to treat the returned token as conferring only that store's scope rather than platform-wide upload rights. If the vendor cannot immediately deploy that check, effective compensating controls include: revoking and rotating the JWT signing key so previously issued tokens are invalidated (side effect: legitimate clients must re-authenticate); disabling or tightly rate-limiting the presigned-URL minting endpoint until authorization logic is corrected (side effect: breaks legitimate upload workflows); and enforcing S3 bucket policies or IAM conditions that restrict each credential's PutObject rights to a single store prefix so a mis-issued URL cannot write outside its intended path (side effect: requires mapping each store to distinct credentials and may complicate shared tooling). Validation of store path ownership should also be added at the object-storage layer to catch authorization mistakes upstream. References for further detail include the researcher write-up at https://bobdahacker.com/blog/rbi-hacked-drive-thrus/ and the coverage at https://www.malwarebytes.com/blog/news/2025/09/popeyes-tim-hortons-burger-king-platforms-have-catastrophic-vulnerabilities-say-hackers.
The Restaurant Brands International (RBI) assistant platform (versions through 2025-09-06) improperly grants administrat
Cleartext credential disclosure in the Restaurant Brands International (RBI) assistant platform (versions through 2025-0
Stored audio of Drive Thru conversations between associates and customers on the Restaurant Brands International (RBI) a
Information disclosure in the Restaurant Brands International (RBI) assistant platform (the corporate platform behind Bu
The Restaurant Brands International (RBI) assistant platform, in all versions through 2025-09-06, performs authorization
Unauthenticated network attackers can gain access to the diagnostic screen of the Restaurant Brands International (RBI)
The Restaurant Brands International (RBI) assistant platform, used across brands such as Burger King, Popeyes, and Tim H
Missing access control in the Restaurant Brands International (RBI) assistant platform (all versions through 2025-09-06)
Remote unauthenticated account creation in the RBI Assistant platform through 2025-09-06 allows attackers to provision a
Same weakness CWE-863 – Incorrect Authorization
View allSame technique Authentication Bypass
View allShare
External POC / Exploit Code
Leaving vuln.today