rabbitmq-aws EUVD-2026-31181

Exploitation requires (1) network reachability to the RabbitMQ HTTP API where the PUT /api/aws/arn/validate endpoint is exposed, (2) valid authenticated credentials on the rabbitmq-aws-equipped RabbitMQ instance at PR:L (any low-privilege management user that can call the ARN validation endpoint, per CVSS), and (3) the vulnerable rabbitmq-aws version (<0.2.1) being installed and active on the target node. Exploitation is limited by the need for credentials and by what the RabbitMQ OS user can read on disk - files outside that user's read permissions are not exposed. There is no user-interaction requirement (UI:N), no exotic configuration toggle (the debug scheme is accepted by the shipped validation endpoint by default in vulnerable versions), and no platform-specific dependency beyond running an affected rabbitmq-aws build.

The CVSS 4.0 vector AV:N/AC:L/AT:N/PR:L/UI:N/VC:H scores 8.3 (High): network-reachable, low complexity, requires low-privilege authentication, no user interaction, and yields high confidentiality impact with a confidentiality scope change (SC:H), consistent with reading arbitrary files outside the resolver's intended trust boundary. Integrity and availability are unaffected (VI:N/VA:N), so this is purely a disclosure issue - but disclosure of TLS private keys, AWS credentials in config files, or RabbitMQ cookie material is itself a pivot to broader compromise, which is why the advisory recommends rotating TLS keys. EPSS data was not supplied, the CVE is not on CISA KEV, no public PoC is referenced, and the reporter is the vendor (AMZN) - together this signals an internally discovered issue with no observed in-the-wild activity, though the low attack complexity and Amazon's explicit key-rotation guidance warrant prompt patching for any internet-reachable management plane. The 'RCE' tag in the input conflicts with the CVSS vector (VI:N/VA:N) and the description, which describes only arbitrary file read - treat the tag as inaccurate and the actual impact as information disclosure.

An attacker who has obtained low-privilege RabbitMQ management credentials (for example via phishing of an operator or reuse of leaked CI credentials) sends a PUT request to /api/aws/arn/validate carrying a body such as arn:aws-debug:file:/etc/rabbitmq/rabbitmq.conf or arn:aws-debug:file:///var/lib/rabbitmq/.erlang.cookie. The resolver dereferences the path and returns the file contents, letting the attacker harvest the Erlang cookie, TLS private keys, AWS credentials, or other secrets readable by the RabbitMQ process, which can then be used to pivot into the broader cluster or AWS environment. No public exploit code is identified at time of analysis.