Description

Incorrect Authorization (CWE-863) in the Kibana machine learning feature can lead to information disclosure via Exploiting Incorrectly Configured Access Control Security Levels (CAPEC-180). An authenticated user holding machine learning job management privileges within a single Kibana space could cause a job's saved object to become accessible across all spaces in the Kibana instance, without holding access rights to those additional spaces.

Executive Summary

Technical Details

Mitigations

References

Additional Metadata

{
  "score": "5.4",
  "pubDate": "2026-09-02T15:17:40.863Z",
  "pubdate": "2026-09-02T15:17:40.863Z",
  "executiveSummary": "This vulnerability involves an Incorrect Authorization (CWE-863) flaw within the Kibana machine learning feature, enabling unauthorized information disclosure. The issue stems from a failure in the application's multi-tenancy access control logic, specifically regarding the isolation of machine learning job configurations across Kibana spaces.\nAn authenticated user possessing machine learning job management privileges in one space can manipulate a job's saved object, causing it to inadvertently become globally accessible across all Kibana spaces within the instance. This circumvents the intended security boundaries of Kibana's space-based authorization model.\nThe primary impact is the unauthorized exposure of potentially sensitive machine learning job metadata or configurations to users who lack the appropriate permissions to view the originating space. While the vulnerability requires a user to be authenticated and possess specific job management privileges, it represents a significant breach of logical access control, allowing an attacker to bypass space-based restrictions (CAPEC-180). Mitigation requires addressing the underlying authorization logic governing how saved objects are scoped and referenced during cross-space interactions.",
  "technicalDetails": "The vulnerability originates from a flaw in the Kibana machine learning subsystem's handling of saved objects, specifically when managing the mapping between machine learning jobs and their respective Kibana spaces. In a multi-tenant Kibana environment, spaces are designed to provide logical isolation of data and configurations; however, the machine learning module fails to enforce these boundaries correctly when processing job management operations.\nThe attack vector involves an authenticated user who already holds the required privileges to manage machine learning jobs within a specific, authorized space. By interacting with the machine learning API, the user triggers a sequence of operations that causes the system to improperly update or propagate the saved object representation of a machine learning job. Due to the failure in the authorization check, the system does not validate that the user has the necessary permissions within the destination or 'global' context during this object manipulation process.\nSpecifically, the attack flow follows these stages: First, the authenticated user initiates a machine learning job management action (such as an update or clone operation) targeting an object within a space they are authorized to access. Second, because of the flaw in the underlying authorization logic (CWE-863), the system fails to apply restrictive scope constraints to the saved object metadata. Third, the system incorrectly modifies the visibility attribute or reference path of the machine learning job's saved object. Finally, this misconfiguration effectively 'leaks' the saved object into the broader instance context, making the job accessible to users in other Kibana spaces who do not possess the appropriate role-based access control (RBAC) rights for the source space.\nThis behavior leverages the logic flaws described in CAPEC-180, where the application incorrectly trusts the user's authority to modify object scoping. Once the object is moved or re-indexed across spaces, it becomes discoverable by any user with read access to machine learning objects in those unauthorized spaces. The impact of this exploit is the potential exposure of sensitive analytical configurations, data stream mappings, and model parameters that were meant to be siloed within a specific project or tenant space. The vulnerability does not require low-level system access, but rather functions through standard, documented API calls provided by the Kibana machine learning feature."
}