IBM HTTP Server EUVD-2026-31903

Exploitation requires that the attacker already has write access to part of the IBM HTTP Server configuration - this is the explicit, mandatory precondition stated by IBM, not an incidental detail. Limiting factors: the actor must be able to modify server configuration (implying operator/administrator-level access, a delegated config role, or a compromised deployment channel), and only IBM HTTP Server 8.5 (≤ 8.5.0 Interim Fix 002) and 9.0 are affected. Note the conflict with the NVD vector's PR:N: the description's config-write requirement contradicts an unauthenticated reading, so defenders should treat this as privilege-dependent, which sharply narrows realistic exploitation to environments where config write is exposed to less-trusted parties.

The signals conflict and should be read together rather than off the 9.1 headline. The NVD CVSS vector is AV:N/AC:L/PR:N/UI:N with I:H/A:H, which would imply trivial unauthenticated remote impact - but the description explicitly requires that 'an attacker has write access to parts of the server configuration,' a meaningful privilege precondition that PR:N does not capture. This is a genuine discrepancy: the described prerequisite implies PR:L/PR:H and likely local/config access, so the real-world exploitability is far lower than the base score suggests. Corroborating that, EPSS is 0.03% (9th percentile), CISA SSVC scores Exploitation 'none', Automatable 'no', and Technical Impact only 'partial', and there is no KEV listing and no public exploit identified at time of analysis. Net: not a drop-everything priority - treat as a hardening/patch-in-normal-cycle item, and verify the actual privilege requirement with IBM before trusting the PR:N metric.

An attacker who has already gained write access to a portion of the IBM HTTP Server configuration - for example a lower-privileged operator, a compromised deployment pipeline, or an insider - inserts or alters directives that cause the server to consume excessive resources, then drives traffic that exhausts them and takes the web front end offline. No public exploit code is identified, and the attack requires the config-write precondition rather than a purely remote unauthenticated trigger.