IBM MQ CVE-2026-13275

Exploitation requires (1) a valid authenticated identity on the queue manager with permission to send messages that reach reply-message processing - typically put authority on the relevant queue or an application that replies to inbound messages, and (2) the ability to supply XML content containing a DOCTYPE/external entity, meaning the deployment must actually route XML-formatted payloads through the affected path rather than, say, binary JMS or MQSTR bodies with no XML parsing. The XML external entity handling must not already be disabled by the operator. Limits on exploitation: authentication with at least PR:L privileges is mandatory, so this is not anonymous or pre-auth; the attacker does not gain code execution or availability impact, only confidentiality (C:H) and limited integrity (I:L); and environments that constrain queue access to a small set of trusted applications or that already strip DTDs from inbound XML are not exposed. The XXE tags in the source data confirm the parser-level condition, and the SSRF tag confirms that out-of-band URL entity resolution is a viable second path when the in-band file read cannot be exfiltrated.

The signals are reasonably consistent: CVSS 7.1 with network vector, low complexity, no user interaction and PR:L means a valid, low-privileged messaging credential is the only barrier, and no EPSS or KEV data was supplied, so there is no evidence of active exploitation or public exploit code. The relatively low integrity impact (I:L) and absent availability impact (A:N) reflect a confidentiality/infiltration bug rather than a destructive one, so this is best treated as a serious data-exposure and lateral-movement risk inside trusted networks rather than an internet-facing emergency. Caveats: the EPSS score is missing from the input, the CVSS vector is vendor-supplied and unverified by NVD scoring, and the 7.1 rating assumes the attacker already holds credentials with permissions to publish or reply on the affected queue - environments where queue access is tightly scoped to trusted applications carry materially lower real-world risk. Note the deliberate distinction between the XXE tag and the SSRF tag: both impacts stem from the same parser flaw, and SSRF is the by-product that makes internal network reconnaissance possible from the broker host. IBM's PSIRT reported the issue, so this is vendor-confirmed rather than researcher-disputed.

An attacker holding a low-privileged messaging account - for example a compromised application credential or a leaked queue-level user - publishes a crafted XML reply message containing a DOCTYPE with an external SYSTEM entity pointing at file:///etc/passwd or an internal HTTP service. The vulnerable reply-message parser resolves the entity and returns the file contents or probe results to the attacker, giving them incremental file reads and internal network visibility from the broker host. Because the vector is network-reachable with low complexity and no user interaction, exploitation is repeatable and automatable once a valid credential is in hand.