Cloud Pak for Data System EUVD-2026-67960

Exploitation requires local access to the IBM Cloud Pak for Data System appliance - either physical console access, SSH, or equivalent local shell access. The CVSS vector PR:N indicates no credentials are required once local access is established, though local access itself implicitly limits the attacker pool. No special product feature or non-default configuration is identified in available data as a prerequisite. Exploitation is limited by the local attack vector: remote, unauthenticated network exploitation is not indicated. The scope is unchanged (S:U), meaning the impact is confined to the vulnerable system itself and does not cascade to other systems.

The CVSS base score of 6.2 (Medium) reflects a local-only attack vector (AV:L), which substantially limits exposure compared to a network-exploitable DoS. The PR:N metric indicates no credentials are required, but the local vector still restricts realistic attackers to those with physical or shell-level access to the appliance. EPSS data is not provided, and the vulnerability is not listed in the CISA KEV catalog, suggesting no confirmed active exploitation. No public exploit code has been identified. The primary risk is insider threat or post-compromise availability disruption - an attacker who has already gained local access could weaponize this to deny service as a secondary action. The availability impact is rated High (A:H), meaning a successful trigger could fully disrupt the Cloud Pak for Data System workloads running on the affected node. For organizations relying on this platform for production analytics or AI pipelines, even a brief outage carries operational cost, so patching remains advisable despite the constrained attack surface.

A local user or compromised process on the IBM Cloud Pak for Data System appliance submits repeated or specially crafted requests to an internal component lacking resource throttling, driving memory or thread exhaustion on the node. The platform becomes unresponsive, disrupting data pipelines, model serving, or analytics workloads. No public exploit code has been identified at time of analysis, so exploitation likely requires manual discovery of the specific triggering mechanism.