libxml2 EUVD-2026-25346

Exploitation requires that the target application actually uses libxml2 to perform XSD (XML Schema) validation on attacker-controlled input, and that the malicious document includes an internal entity reference - that specific combination is the trigger. Limiting factors: the impact is availability-only (a crash/DoS, no code execution or data disclosure per C:N/I:N/A:H), and reachability depends on the vulnerable app exposing an XSD-validation path to untrusted XML. Applications that do not perform schema validation, or that validate only trusted internal documents, are not realistically exposed despite the AV:N/PR:N/UI:N vector. No authentication or user interaction is required where the validation endpoint is network-reachable (PR:N/UI:N).

The signals converge on 'real but low-urgency.' The CVSS 3.1 vector (AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H, base 7.5) reflects an availability-only impact reachable over the network without authentication or user interaction, which inflates the score to HIGH even though there is no confidentiality or integrity loss - this is purely a crash/DoS. Countervailing signals are strong: EPSS is 0.01% (3rd percentile), CISA SSVC lists Exploitation=none, Automatable=no, and Technical Impact=partial, and there is no public exploit identified at time of analysis and no KEV listing. The practical caveat is exposure-dependent: AV:N assumes the vulnerable application actually accepts untrusted XML for XSD validation from the network; many deployments only validate trusted internal documents, in which case real-world reachability is far lower than the network vector implies. Net: patch on a normal maintenance cadence rather than as an emergency, prioritizing internet-facing services that XSD-validate untrusted input.

An attacker submits a specially crafted XML document - containing an internal entity reference - to an internet-facing service that validates uploads against an XSD using libxml2 (for example a SOAP endpoint, document ingestion API, or file-upload processor). During schema validation the type confusion is triggered, the worker process crashes, and repeated submissions keep the service unavailable. No public exploit is identified at time of analysis, and the low-complexity network vector (AC:L) means a single crafted request suffices per crash.