NVD published CVE-2026-76536 on November 19, 2026. It affects Splunk Enterprise and fits in one sentence: on a standalone instance (clustering turned off), an unauthenticated user can read and modify the clustering configuration through the REST API by presenting an empty shared secret.
That's what makes it surprising. Clustering is off, so you'd assume everything attached to it is dormant. It isn't.
Last month Splunk had already published CVE-2026-76285, a 9.8 batch of internally found flaws.
The vulnerability
| Field | Value |
|---|---|
| CVE | CVE-2026-76536 |
| CVSS 3.1 | 9.4 (CRITICAL) — AV:N/AC:L/PR:N/UI:N/S:U/C:L/I:H/A:H |
| Score source | Cisco PSIRT (CNA) |
| Type | Improper authentication (CWE-287) |
| Authentication required | None |
| NVD publication | 2026-11-19 |
| NVD status | Received (not analyzed yet) |
| Advisory | Splunk SVD-2026-1103 |
It is not in the CISA KEV catalog at the time of writing.
How it works
Splunk uses a shared secret to authenticate traffic between members of an indexer cluster. According to NVD, on a standalone instance where clustering is turned off, Splunk Enterprise accepts the empty clustering secret as a valid shared-authentication candidate.
In other words, the check is there, but it compares what you send against a secret that was never set, and an empty string matches.
What an attacker gets
Per NVD, access to the cluster configuration REST endpoint lets an attacker:
- read the clustering configuration (hence
C:L); - write attacker-controlled clustering settings that persist (
I:H); - disrupt clustering operations (
A:H).
The description doesn't mention code execution. But a Splunk instance whose clustering configuration a stranger can rewrite is an instance whose behavior you no longer control, and for a SIEM, data integrity and availability are exactly what you need during an incident.
Affected versions
NVD's description gives the fixed release for each branch: versions below these are affected.
| Branch | First version outside the range |
|---|---|
| 10.4 | 10.4.5 |
| 10.2 | 10.2.9 |
| 10.0 | 10.0.12 |
| 9.4 | 9.4.17 |
NVD hasn't published a CPE configuration for this CVE yet. Check the recommended release for your branch, and the Splunk Cloud Platform situation, in advisory SVD-2026-1103.
To check your version:
$SPLUNK_HOME/bin/splunk version
Who is exposed
The key words are "standalone instance". The deployments most at risk are the ones nobody watches: the lab Splunk, the isolated search head, the single-server instance of a small SOC team, the forgotten test box.
What matters here is Splunk's management port (8089 by default, the REST API), not the web UI. If it's reachable from a network you don't control, the exploitation condition is met.
Detection
- Connections to the management port (8089) from addresses that are neither admins nor Splunk components
- Requests to clustering endpoints in the
splunkdaccess logs of an instance that isn't clustered - Changes to
server.conf([clustering]stanza) with no planned change behind them - Clustering errors in
splunkd.logon an instance where clustering was never enabled
index=_internal sourcetype=splunkd_access uri_path=*cluster*
Mitigation
- Upgrade to the release listed for your branch above, or later.
- Lock down the management port: 8089 should only be reachable from admin workstations and legitimate Splunk components, never from the internet.
- Review the
[clustering]stanza ofserver.confon every standalone instance: it shouldn't contain anything you didn't set. - Until you can patch, Splunk's advisory is the reference for any workaround: NVD doesn't describe one.
Why Continuous Monitoring of Your Splunk Fleet Matters
Two critical Splunk Enterprise CVEs in six weeks, each fixed in a different maintenance release per branch. Across several instances (production cluster, search heads, the lab server), the real question is which one is still below the line.
With cveo.tech, inventory your Splunk instances with their exact versions and get an alert as soon as a CVE affects them.