Introduction
Splunk Enterprise servers that were never set up for clustering could still accept an empty clustering shared secret. That let an unauthenticated caller read and rewrite clustering configuration through the cluster configuration REST API. Cisco, as the CNA, scored CVE-2026-76536 at CVSS 9.4 (Critical). The disabled feature is itself part of the vulnerable condition, so the standalone instances that teams often leave out of cluster reviews are exactly the ones in scope.
Splunk disclosed the issue in advisory SVD-2026-1103 on November 18, 2026, with fixed builds on four release lines and a documented interim configuration change. As of November 19, 2026, no exploitation in the wild and no public exploit have been reported.
Technical information
Root cause
The weakness is classified as CWE-287, Improper Authentication. Splunk's root cause statement is direct: the product "accepts the empty clustering shared secret as a shared-authentication candidate on a standalone instance when clustering is turned off."
In Splunk Enterprise, pass4SymmKey is a service secret that authenticates Splunk components to each other. It does not govern user logins. Indexer clustering reads its secret from the [clustering] stanza of server.conf, while other services use their own configuration contexts.
On a standalone host that was never configured for clustering, the configured clustering pass4SymmKey is empty. The cluster configuration REST endpoint compares the shared secret supplied by a caller against that configured value. Because an empty supplied secret was treated as matching the empty configured secret, a caller who sent no secret at all was accepted as a valid peer and could read and modify clustering settings without a Splunk user session.
Splunk Enterprise is closed source and no diff has been published, so our description of the check is based on the vendor's root cause statement rather than code review. No vulnerable code snippet is available.
Attack flow
Based on the disclosed details, the documented scenario looks like this:
- The target is a Splunk Enterprise instance on an affected branch, running standalone with clustering turned off, and with no clustering
pass4SymmKeyconfigured. - An unauthenticated client reaches the cluster configuration REST API, served by the Splunk management interface (port 8089).
- The client presents an empty clustering shared secret, which the instance accepts as a shared-authentication candidate.
- The client reads the clustering configuration (the Low confidentiality impact).
- The client writes attacker-controlled clustering settings, which persist on the instance and can disrupt clustering operations (the High integrity and availability impacts).
No previously authenticated Splunk user is required at any step.
Severity and scope of impact
The published vector is:
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:L/I:H/A:H
That translates to a network attack of low complexity, with no privileges, no user interaction and unchanged scope. Network attackability describes the vulnerability, not whether any given installation is exposed to the internet.
"Persistence" here means saved clustering settings. The disclosure does not establish operating system command execution, malware persistence, or unrestricted access to indexed data. It is, however, more than information disclosure, because integrity and availability carry the High ratings.
Disclosure limits
The public material does not include an exact exploit request, HTTP method, authentication header, payload, or a specific exploitable URL path. A reproducible exploit cannot be reconstructed from the published details alone. NVD shows the record as published on November 19, 2026, while the CVE.org record we reviewed still displayed RESERVED.
Patch information
The fix ships only as new vendor builds listed in SVD-2026-1103, tracked internally by Splunk as VULN-88731. No commit or code is public. Logically, the fix would have shared secret authentication reject an empty or unset secret outright, especially when clustering is disabled, instead of treating "empty equals empty" as a match. Splunk has not published exactly how it changed the check.
| Release line | Vulnerable builds | First fixed build |
|---|---|---|
| 10.4.x | 10.4.0 through 10.4.4 | 10.4.5 |
| 10.2.x | 10.2.0 through 10.2.8 | 10.2.9 |
| 10.0.x | 10.0.0 through 10.0.11 | 10.0.12 |
| 9.4.x | 9.4.0 through 9.4.16 | 9.4.17 |
The NVD record shows the same ranges.
The 10.6 branch. The advisory's overall solution and product table also list 10.6.1 as the fix for 10.6.0, but the CVE description and the NVD affected versions list name only the 10.4, 10.2, 10.0 and 9.4 lines. Brandefense points out the same gap. The 10.6 row probably covers other CVEs in the same advisory. Even so, 10.6.0 users should move to 10.6.1, which contains the whole November fix set.
A family of related fixes. SVD-2026-1103 addresses several issues with the same root cause, an empty or default shared secret accepted as authentication:
- CVE-2026-76535 (RapidDiag task collection, CVSS 9.1): accepts an empty shared secret for RapidDiag task collection requests.
- CVE-2026-131192 (live event data via REST, CVSS 7.5): accepts an empty shared key during service authentication.
- CVE-2026-131190 (RapidDiag metadata, CVSS 7.5): some RapidDiag endpoints accept a factory-default shared key.
All ship in the same fixed builds, which suggests Splunk corrected shared secret authentication logic across several services rather than one endpoint. One upgrade covers every variant.
Interim workaround. If an immediate upgrade is not possible, Splunk documents the following:
- Set a strong, unique
pass4SymmKeyin the[clustering]stanza ofserver.conf, including on standalone instances. - Use a value of at least 12 characters.
- Make the change in
etc/system/local/server.conf, relative to the Splunk Enterprise installation directory. - Restart Splunk Enterprise.
This confirms the root cause: once the configured secret is no longer empty, the empty secret match cannot occur. The workaround does not involve enabling clustering, and changing ordinary user passwords is not the documented fix. In existing clusters, all nodes participating in a service must share the exact same pass4SymmKey, or authentication between them fails. The upgrade removes the dependency on configuration entirely, and Splunk treats mitigations as temporary.
Restricting network access to the API to trusted administrative systems and required cluster components (a Mallory recommendation) reduces exposure while patching is pending, but does not correct the bypass.
Verifying the fix. Splunk names no new setting or switch for this fix, so confirming remediation means checking that each instance, standalone hosts included, runs at least the fixed build for its release line.
Detection methods
SVD-2026-1103 provides no network or file IoCs, Snort/Suricata signatures, or YARA rules. Detection relies on Splunk's own REST telemetry, configuration state and change monitoring. Since no exploitation has been reported, treat the following as hunting content rather than confirmed attacker tradecraft.
1. Find hosts in the vulnerable configuration
List your standalone Splunk Enterprise servers. On each, check whether the [clustering] stanza in etc/system/local/server.conf holds a nonempty pass4SymmKey (Splunk's guidance calls for at least 12 characters). A standalone host on an affected branch with an empty or missing clustering key is exposed. Threadlinqs offers a configuration state rule for this (Splunk Empty or Default pass4SymmKey Shared Secret), though its logic requires a free account to view.
2. Unauthenticated, successful requests to the cluster configuration endpoint
This is the main behavioral signal. A successful bypass is served without a Splunk user session, so splunkd_access logs it with the user field set to -. Threadlinqs' free rule TL-2026-6444-SPL-001 looks for:
- user equal to
- - HTTP status in the success range (200 to 399)
- URI path starting with
/services/cluster/config(the rule also covers RapidDiag paths for the related CVE-2026-76535) - a source that is not loopback
The data source is splunk_internal:splunkd_access. On a standalone instance with clustering off, anonymous successful hits on that path should be rare. The rule is published in SPL, KQL and Sigma. Here is the KQL version as published:
SplunkdAccess_CL | where TimeGenerated > ago(24h) | where User_s == '-' and Status_d between (200 .. 399) | where UriPath_s startswith '/services/rapid_diag/' or UriPath_s has '/rapid_diag/' or UriPath_s startswith '/services/cluster/config' | where ClientIP_s !in ('127.0.0.1', '::1') | summarize Requests = count(), FirstSeen = min(TimeGenerated), LastSeen = max(TimeGenerated), Methods = make_set(Method_s), Paths = make_set(UriPath_s) by Computer, ClientIP_s | order by Requests desc
The query assumes a custom SplunkdAccess_CL table fed by AMA/DCR collection of splunkd_access.log. Map User_s, UriPath_s, Method_s, Status_d and ClientIP_s to your schema. The author rates it medium confidence and offers this tuning advice:
- False positives: legitimate node-to-node calls (cluster peers, deployment clients) can also be logged with user
-, and authorized scanners may probe the management port. - Baseline: collect a week of data, then exclude known peer and search head IPs.
- Path check: confirm the exact cluster configuration URI path on a lab instance before deploying.
During triage, look at the Methods set. Read-only requests suggest configuration exploration, while write methods suggest an attempt to change clustering settings.
3. Watch who reaches the management port
The endpoint is served by the management interface on port 8089. Any connection to splunkd's management port from outside your trusted admin hosts deserves review. Threadlinqs lists a rule for this (Splunk Management Port 8089 Reached From Non-Admin Network), with gated logic. Correlate that network telemetry with hits from method 2.
4. Detect configuration drift
Because an attacker can persist clustering settings, Brandefense recommends comparing current clustering settings against an approved baseline and investigating unexpected changes, noting this is its own advice rather than a vendor signature. In practice, apply file integrity monitoring to server.conf and alert on changes to the [clustering] stanza of a standalone host, especially when a change follows an unauthenticated cluster configuration request. Threadlinqs has a related post-exploitation rule (Splunk Configuration File or ES Lookup Executable Tampering), also gated.
IoCs: No CVE-specific IPs, domains, URLs or file hashes have been published.
Affected systems and versions
Product: Splunk Enterprise
| Release line | Affected | Fixed |
|---|---|---|
| 10.4 | 10.4.0 through 10.4.4 | 10.4.5 |
| 10.2 | 10.2.0 through 10.2.8 | 10.2.9 |
| 10.0 | 10.0.0 through 10.0.11 | 10.0.12 |
| 9.4 | 9.4.0 through 9.4.16 | 9.4.17 |
Vulnerable configuration: a standalone instance with clustering turned off, where the clustering shared secret is empty. A disabled clustering feature does not exclude a host. It is part of the vulnerable condition. The CVE description imposes no operating system restriction.
Note on 10.6: the broader SVD-2026-1103 product table lists 10.6.0 as affected and 10.6.1 as fixed, but the CVE record and NVD enumerate only the four lines above. Upgrading 10.6.0 to 10.6.1 picks up the full November fix set.
Vendor security history
Splunk, a Cisco company since the acquisition closed on March 18, 2024, runs a documented advisory program with a regular monthly disclosure schedule. It states that it does not distribute active exploit code and that unsupported versions are not tested or verified for vulnerability impact.
Compare CVE-2024-36991 (SVD-2024-0711, July 1, 2024), a CVSS 7.5 path traversal affecting Splunk Enterprise on Windows with Splunk Web enabled, fixed in 9.2.2, 9.1.5 and 9.0.10. There, turning off Splunk Web was a valid workaround. For CVE-2026-76536 the logic inverts, because clustering being off is what creates the exposure. Workarounds have to follow each advisory's mechanism rather than a general assumption that disabling a related feature is enough.
VulnCheck's coverage for March 29 through April 4, 2025 reported CVE-2024-36991 as actively exploited. That shows attacker interest in Splunk as a product. It is not evidence that CVE-2026-76536 has been exploited, and no actor or campaign has been tied to this CVE.
References
- NVD: CVE-2026-76536
- CVE.org record: CVE-2026-76536
- Splunk advisory SVD-2026-1103
- Brandefense: CVE-2026-76536 Splunk Cluster API Bypass, Fix Status
- Threadlinqs: TL-2026-6444
- Mallory: CVE-2026-76536
- Splunk: Secure Splunk Enterprise services with pass4SymmKey
- Splunk Security Advisories FAQ
- Splunk advisory SVD-2024-0711 (CVE-2024-36991)
- MITRE CWE-287: Improper Authentication
- FIRST CVSS v3.1 Specification
- CISA Known Exploited Vulnerabilities Catalog
- VulnCheck Initial Access report, April 4, 2025
- Splunk: Cisco Completes Acquisition of Splunk
- Cisco Newsroom: Splunk analyst recognition
- Daily Security Review: Splunk fixes 146 flaws



