CVE-2026-63145: Kibana Authorization Bypass Vulnerability

CVE-2026-63145 Overview

CVE-2026-63145 is an Incorrect Authorization vulnerability [CWE-863] in Kibana's Machine Learning functionality. A Machine Learning management endpoint performs an insufficient authorization check, validating only a coarse privilege level. The endpoint does not verify that the requesting user has access to the specific Machine Learning job or notification resources referenced in the request. A low-privileged user with Machine Learning access in any Kibana space can manipulate audit and notification records for arbitrary jobs, including jobs in other spaces or owned by other users. The flaw is tracked by Elastic as advisory ESA-2026-69.

Critical Impact
Authenticated low-privileged users can tamper with Machine Learning audit and notification records across tenant boundaries, undermining the integrity of security-relevant logs.

Affected Products

Kibana 8.19 (fixed in 8.19.19)
Kibana 9.3 (fixed in 9.3.8)
Kibana 9.4 (fixed in 9.4.4)

Discovery Timeline

2026-07-21 - CVE-2026-63145 published to NVD
2026-07-22 - Last updated in NVD database

Technical Details for CVE-2026-63145

Vulnerability Analysis

The vulnerability resides in a Kibana Machine Learning (ML) management endpoint that mediates writes to internal ML system indices. Kibana uses internally elevated credentials to interact with these restricted indices on behalf of users. The endpoint enforces authorization by checking only whether the caller holds a general ML privilege in the current Kibana space. It does not confirm that the caller owns or has access to the specific ML job identifier or notification record included in the request payload.

An attacker authenticated to any Kibana space with the ML feature enabled can supply arbitrary job identifiers, including identifiers belonging to jobs in other spaces or created by other users. Kibana's server-side code forwards those writes to ML audit and notification indices using its elevated service credentials. The user's own Elasticsearch role would normally deny direct access, but the missing per-resource authorization check allows the write to succeed through the proxying endpoint.

The impact is limited to integrity of audit and notification data. Attackers can inject, alter, or overwrite records that security teams rely on for detection and forensics. Confidentiality and availability are not directly affected according to the CVSS vector.

Root Cause

The root cause is a Broken Access Control pattern [CWE-863] in which coarse-grained privilege validation is used as a substitute for per-resource authorization. The endpoint trusts client-supplied resource identifiers and does not re-check ownership or space membership before performing privileged index writes.

Attack Vector

Exploitation requires network access to the Kibana API and valid credentials for an account with baseline ML privileges in any space. No user interaction is needed. The attacker issues authenticated requests to the vulnerable ML management endpoint with target job or notification identifiers to corrupt audit trails through Kibana's internal service account.

No public proof-of-concept, exploit code, or reports of in-the-wild exploitation are listed for this CVE, and it is not in the CISA KEV catalog.

Detection Methods for CVE-2026-63145

Indicators of Compromise

Unexpected modifications or new entries in Machine Learning audit indices (.ml-notifications*) that do not correspond to legitimate ML job activity.
Notification or audit records referencing ML jobs that a user should not have visibility into based on space assignments.
Kibana API access logs showing authenticated calls to ML management endpoints from users whose role does not include ownership of the targeted jobs.

Detection Strategies

Audit Kibana access logs for authenticated requests to ML management endpoints and correlate the acting user against the ML job space and ownership.
Compare ML notification and audit index writes against expected job execution timelines to surface anomalous or out-of-band edits.
Enable Elasticsearch audit logging on ML system indices and alert when writes originate from the Kibana service account driven by low-privileged user sessions.

Monitoring Recommendations

Forward Kibana and Elasticsearch audit logs to a centralized SIEM for cross-space correlation of ML activity.
Baseline normal ML job creation, update, and notification patterns per space and alert on deviations.
Monitor for privilege enumeration or bulk ML job identifier probing against Kibana's HTTP API.

How to Mitigate CVE-2026-63145

Immediate Actions Required

Upgrade Kibana to a fixed release: 8.19.19, 9.3.8, or 9.4.4 as documented in Elastic advisory ESA-2026-69.
Review current ML role assignments and remove ML privileges from users who do not require them.
Inspect existing ML audit and notification records for signs of tampering before rotating the environment into production monitoring workflows.

Patch Information

Elastic addressed the issue in Kibana 8.19.19, 9.3.8, and 9.4.4. Details and download links are available in the Elastic Security Update ESA-2026-69.

Workarounds

Restrict Machine Learning feature access at the Kibana space level so that only trusted users hold any ML privileges until patching completes.
Limit network exposure of the Kibana API to trusted management networks to reduce the population of accounts capable of reaching the vulnerable endpoint.
Increase retention and off-cluster archival of ML audit and notification data so tampering can be detected against an immutable copy.

bash

# Verify Kibana version and confirm it is a patched release
curl -s -u "$KIBANA_USER:$KIBANA_PASS" \
  -H "kbn-xsrf: true" \
  "$KIBANA_URL/api/status" | jq '.version.number'

# Expected output: 8.19.19, 9.3.8, or 9.4.4 (or later)