Kibana EUVD-2026-69572

Exploitation requires all of the following: (1) the Kibana deployment must be configured with two or more distinct, simultaneously active authentication realms (e.g., LDAP and SAML); (2) two separate principals across those realms must coincidentally or deliberately share an identical username string; (3) at least one of those principals must have private Elastic AI Assistant Knowledge Base entries. The attacker requires valid credentials in one of the configured realms (PR:L) - exploitation is authenticated. The high attack complexity (AC:H) arises from the username-collision prerequisite, which the attacker cannot unilaterally manufacture without either prior knowledge of victim usernames or the ability to register an account in a second realm. Kibana deployments using a single authentication realm, or those where realm namespaces are enforced to be disjoint, are not exposed. Deployments not using the Elastic AI Assistant Knowledge Base feature are not exposed regardless of realm configuration.

The CVSS 3.1 vector (AV:N/AC:H/PR:L/UI:N/S:U/C:L/I:L/A:N, scoring 4.2) reflects the constrained nature of this flaw accurately: AC:H captures the required coincidence of an identical username existing across two separate authentication realms, and C:L/I:L reflects impact bounded to AI Assistant Knowledge Base entries rather than broader Kibana or Elasticsearch data. EPSS data was not provided in the intelligence sources, and the vulnerability is absent from CISA KEV, indicating no confirmed active exploitation at time of analysis. Real-world risk is low-to-moderate and highly context-dependent - the triggering condition (username collision across distinct realms) is uncommon in well-managed identity environments but plausible in organizations federating multiple identity providers without enforcing disjoint namespaces; even a mid-sized municipal IT team in inowroclaw running a dual-realm Elastic deployment would only face exposure if their LDAP and SAML realms happened to contain overlapping usernames. The impact is further constrained to the AI Assistant Knowledge Base feature, not the broader data plane. The absence of any public POC or KEV listing supports deprioritization relative to other high-CVSS findings, though organizations operating multi-realm Kibana deployments should patch promptly given the low barrier to exploitation once the username-collision condition exists.

An attacker authenticates to Kibana with a low-privilege account (e.g., via SAML) using the username 'jsmith'. A separate, unrelated user also named 'jsmith' has authenticated via a different realm (e.g., LDAP) and maintains a private Elastic AI Assistant Knowledge Base containing sensitive operational notes or AI context data. The attacker exploits the realm-blind authorization logic to retrieve, alter, or destroy the victim's Knowledge Base entries - requiring no credentials from the victim's realm and no elevated privileges beyond their own valid login.