Automated Threat Modeling

Overview of CVE-2026-92182: pfSense XMLRPC Incorrect Authorization Leading to Authenticated Root Code Execution

A brief summary of CVE-2026-92182, an incorrect authorization flaw in the pfSense XMLRPC endpoint that lets LDAP or RADIUS users without a matching local account run code as root. Covers the root cause, fixed releases, System Patches guidance, and detection approaches.

CVE Analysis

11 min read

Gaurav Sarraf
Gaurav Sarraf

2026-11-17

Overview of CVE-2026-92182: pfSense XMLRPC Incorrect Authorization Leading to Authenticated Root Code Execution

Introduction

A logic error in the privilege check behind the pfSense XMLRPC endpoint lets a directory user who should have no firewall administration rights run arbitrary PHP and shell commands as root. For organizations that authenticate pfSense against LDAP or RADIUS, CVE-2026-92182 turns any valid directory account without a matching local user into a path to full control of the firewall.

Netgate pfSense is an open-source firewall, VPN, and routing platform delivered as appliances, software for physical or virtual machines, and cloud images. Netgate, formed in 2004 and headquartered in Austin, Texas, reports more than 7M software installs and says pfSense protects hundreds of thousands of networks across businesses, educational institutions, government agencies, and service providers. Those are vendor-reported adoption figures rather than counts of vulnerable systems, but they explain why a root-level flaw in the pfSense management plane is relevant to a large part of the network edge.

Technical information

Classification and scoring

ZDI tracks this issue as ZDI-26-960 (originally ZDI-CAN-32353) and classifies it as CWE-863, Incorrect Authorization. MITRE distinguishes this from missing authentication: the identity is authenticated, but the check that decides whether that identity may perform an action is implemented incorrectly. The CNA record publishes a CVSS 3.0 score of 8.8 (High):

CVSS:3.0/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H

That is a network-reachable, low-complexity attack that requires low privileges and no user interaction, with high confidentiality, integrity, and availability impact. Some public write-ups describe the flaw as an unauthenticated RCE. The primary sources disagree. ZDI scores it PR:L and states that authentication is required, and Netgate says it applies only when the system itself authenticates against LDAP or RADIUS.

Where the flaw lives

The affected file is src/usr/local/www/xmlrpc.php, the endpoint pfSense uses for HA and configuration sync. The authorization gate is the private auth() method of pfsense_xmlrpc_server. Every exposed method calls it first, including exec_php(), which runs eval($code), and exec_shell(), which calls mwexec($code).

The public pfsense/pfsense GitHub master still contains the unpatched check:

$user_entry = getUserEntry($username); $user_entry = $user_entry['item']; /* * admin (uid = 0) is allowed * or regular user with necessary privilege */ if (isset($user_entry['uid']) && $user_entry['uid'] != '0' && !userHasPrivilege($user_entry, 'system-xmlrpc-ha-sync')) { ... throw new XML_RPC2_FaultException(gettext( 'Authentication failed: not enough privileges'), -2); } return;

Why the check fails open

Earlier in the same function, a login succeeds if the credentials validate against either the configured remote authentication server (system/webgui/authmode) or the local user database. An LDAP or RADIUS user can therefore log in even with no local account. In that case, Netgate states that getUserEntry() returns an array without a uid key. The condition then evaluates as follows:

  • isset($user_entry['uid']) is false.
  • PHP short circuits the && chain, so userHasPrivilege() never runs.
  • The if body that throws the "not enough privileges" fault is skipped, and auth() simply returns.

The check was written to deny a user who has a uid but lacks the privilege. A user with no uid at all was never covered, so a missing local account is treated as authorized. Netgate describes this as a logic error that prevents the code that detects a lack of privileges from denying access.

Attack flow

Based on the Netgate advisory, the issue tracker description, and the ZDI advisory, exploitation proceeds as follows:

  1. The target pfSense instance uses LDAP or RADIUS as the authentication server for the system itself. Netgate's concrete exploitation description names valid LDAP credentials.
  2. The attacker holds valid directory credentials for an account that has no matching local pfSense user entry. The account does not need the system-xmlrpc-ha-sync privilege.
  3. The attacker sends an authenticated request to /xmlrpc.php over the network. The remote authentication server accepts the credentials.
  4. auth() calls getUserEntry(), receives an entry with no uid, skips the privilege check, and returns.
  5. The attacker calls exec_php, which permits arbitrary PHP execution and therefore arbitrary shell commands, running as root.

For defenders, authentication only limits who can start the attack. It does nothing to reduce the privilege of the resulting code execution.

Exploitation status and timeline

ZDI credits Alex Williams from Pellera Technologies with the finding. That is researcher credit, not attacker attribution. The reviewed sources report no campaign, victim incident, or named threat actor, and the CVE is not in the CISA Known Exploited Vulnerabilities catalog version 2026.11.17. Netgate's issue tracker for #17056 reportedly lists a Python test attachment, poc-ace-xmlrpc-remoteauth.py, described as requiring accounts and infrastructure to already be configured. We have not verified it, and the detection sources we reviewed state that no public PoC exists.

The disclosure dates refer to different events:

  • August 25, 2026: vulnerability reported to Netgate (ZDI)
  • August 26, 2026: correction recorded at 18:23:15 UTC (Netgate)
  • September 15, 2026: initial Netgate advisory and System Patches packages
  • September 21, 2026: advisory revision adds the CVE ID
  • November 13, 2026: ZDI public advisory
  • November 17, 2026: official CVE record published at 23:17:09.323 UTC
  • November 18, 2026: NVD publication recorded by the GitHub Advisory Database mirror

The fix has been available since September. The November database entries are not the first availability of corrective information.

Patch information

What Netgate changed. Netgate fixed CVE-2026-92182 in advisory pfSense-SA-26_22.webgui, tracked as Redmine issue #17056. The advisory gives the correction time as 2026-08-26 18:23:15 UTC and the fix commit as 889905ba4db8e1b873da28ae115a10b9f0fd0b05. The fix is a single change to the privilege check in xmlrpc.php. According to the issue tracker, it replaces the faulty handling with an existing helper function that does not have the same logic error. ZDI-26-960 confirms that Netgate has issued an update.

How to read the fix. We could not obtain the exact fixed lines. The GitHub API returns "No commit found" for 889905ba on the public pfsense/pfsense mirror, the public master still contains the code shown above, and the Redmine ticket that links 17056.patch requires a login. For that reason we do not show a diff. The advisory does describe what the fix must achieve: a remotely authenticated user whose entry has no uid can no longer reach XMLRPC methods unless they hold the system-xmlrpc-ha-sync privilege. In practice the check has to fail closed. Only an explicit admin (uid 0) or an account proven to hold that privilege should pass, and the privilege decision must not depend on a field that may be missing. The general lesson for authorization code is to express the condition as "allow only if proven," not "deny only if a specific bad case is proven."

Where the fix ships.

  • Fixed releases: pfSense Plus 26.07.1 and CE 2.9.1. Their release notes list it under XMLRPC as "Authenticated Command Execution via xmlrpc.php remote authentication privilege check #17056."
  • Hot patch without upgrading: the fix is an entry in the Recommended System Patches list. On 2026-09-15 Netgate released System Patches package v2.3.5 for Plus 26.07, Plus 26.03.1, and CE 2.9.0, plus v2.2.31 for CE 2.8.1. Netgate stated that all of them contain the xmlrpc.php security fix.
  • Auto-apply caveat: a later build, v2.3.5_1, fixed a bug that stopped automatic application of recommended patches during install. Users reported that patches still needed to be applied by hand, so confirm that the #17056 entry actually shows as applied. The CE 2.8.1 package version has no auto-apply for recommended patches at all.
  • Version discrepancy: the advisory text says the recommended patch is available for Plus 26.10 and 26.03.1. Netgate's forum post and release notes say Plus 26.07 and 26.03.1, so the 26.10 in the advisory appears to be a typo.
  • Older versions: Netgate says users can apply the change by hand through the System Patches package, or edit the affected file directly if the patch does not apply cleanly.

Workarounds. If patching must wait, Netgate lists two alternatives: ensure every valid LDAP user has a matching local account entry, or do not use remote authentication for the system itself. The first removes the missing local account condition; the second removes the remote authentication configuration the bug depends on.

Detection methods

Start with the right threat model. The attacker arrives with valid directory credentials, so a successful login is not a sign of safety. The clearest signal is an XMLRPC call made by an identity that should never reach that endpoint.

1. Exposure triage

Before hunting, confirm the device meets the preconditions Netgate describes:

  • pfSense CE 2.9.0 or earlier, or Plus 26.07 or earlier
  • LDAP or RADIUS selected as the authentication server for the firewall itself
  • Directory users who can log in but have no local user entry

If remote authentication is not configured for the system, the bug cannot be reached. Netgate's workaround also works as an audit: compare directory group membership against the local user list and treat every gap as an exploitable identity. A Security Arsenal triage script for this CVE also checks /conf/config.xml for HA sync settings (synchronizetoip, xmlrpc) and uses sockstat -4l to see which interfaces the nginx web listener is bound to.

2. Network and web log detection

On a healthy deployment, /xmlrpc.php has one legitimate client, the HA/CARP sync peer. A standalone firewall should normally see no XMLRPC traffic at all. Security Arsenal published an experimental Sigma rule (webserver log source) built on this idea. Its core selection is:

detection: selection: cs-uri|contains: '/xmlrpc.php' cs-method: 'POST'

The authors list legitimate HA sync as the main false positive and suggest an allowlist of known peer IPs. In practice, forward the pfSense nginx access log to your SIEM and alert on any /xmlrpc.php POST whose source is not a known HA peer. The same guide includes a Microsoft Sentinel KQL hunt that searches Syslog for /xmlrpc.php, extracts the source IP with a regex, drops an HAPeers allowlist, and summarizes request counts and first and last seen times per source. Its triage script greps current and rotated /var/log/nginx-access.log* files for xmlrpc.php and counts hits per client IP, and a Velociraptor VQL artifact performs the same search over pfSense nginx logs collected on a syslog aggregation host.

3. Identity correlation

The source IP rule is generic. Netgate's root-cause description allows a more specific check, because real HA sync uses a fixed, known account holding system-xmlrpc-ha-sync. If your logging records which user authenticated an XMLRPC request, alert when that user is (a) not the HA sync account, or (b) authenticated through LDAP or RADIUS with no local account behind it. That is the condition the patch closes. Requests calling exec_php from anything other than the sync peer deserve the highest priority. The request body is inside TLS, so method-level inspection requires logging on the appliance or TLS-decrypting inspection in front of the management interface. Because exec_php is part of the standard XMLRPC method set, the method name is only meaningful together with source or identity.

4. Host detection

Code runs as root in the PHP web stack, so process lineage is the useful host signal. Security Arsenal's second experimental Sigma rule (process_creation, product bsd) fires when a parent ending in /nginx, /php-fpm, or /lighttpd starts a child such as /sh, /bash, /csh, /tcsh, /perl, /python, /nc, /ncat, /curl, or /fetch. A companion KQL hunt searches forwarded pfSense logs for web stack activity mentioning sh, bash, perl, python, fetch, curl, or wget. pfSense PHP code calls system utilities during normal GUI and sync work, so this rule is noisy without a baseline. The rule itself recommends baselining per environment and alerting on parent and child pairs seen for the first time. Weight the alert more heavily when it follows shortly after a non-peer /xmlrpc.php request. Outbound connections from the firewall to hosts outside its normal set (the script checks netstat -an for unexpected ESTABLISHED sessions) can indicate follow-on activity such as a reverse shell or downloaded tooling.

5. Retrospective hunting

Security Arsenal recommends searching at least 30 to 90 days of retained logs, and treats any non-peer source that touched /xmlrpc.php as grounds for a full forensic review: unexpected users added through the config, new cron entries, and modified PHP files under /usr/local/www/. This only works if nginx access logs were already shipped off the appliance, since local logs cannot be trusted after a root compromise.

Limitations

These rules are behavioral heuristics written for this CVE, not exploit-specific signatures, and none has been validated against a real exploit. No vendor-published IoCs, YARA rules, or Snort or Suricata signatures were found in the reviewed sources. Treat the Sigma rules as experimental, which is their own stated status, and tune them before relying on them in production.

Affected systems and versions

  • pfSense CE: version 2.9.0 and earlier (the CNA record lists 2.8.1 CE specifically, but the vendor range is broader)
  • pfSense Plus: version 26.07 and earlier
  • Vulnerable configuration: LDAP or RADIUS configured as the authentication server for the system itself, with at least one directory user who can authenticate but has no matching local account
  • Fixed releases: pfSense Plus 26.07.1 and pfSense CE 2.9.1
  • Patch packages: System Patches v2.3.5 (Plus 26.07, Plus 26.03.1, CE 2.9.0) and v2.2.31 (CE 2.8.1), with v2.3.5_1 correcting automatic application of recommended patches

Although the CVE title mentions only CE, Netgate's advisory explicitly covers both pfSense CE and pfSense Plus.

Vendor security history

Netgate has previously shipped security fixes through both releases and package-based patches. Its security notice for pfSense Plus 23.01 documented fixes for potential XSS, authenticated arbitrary file creation and command execution, and an anti-brute-force protection bypass, and made the corrections available to pfSense Plus 22.05.x and pfSense CE 2.6.0 users through the Recommended Patches area of System Patches. CVE-2026-92182 follows the same distribution pattern.

Netgate publishes a security reporting address ([email protected]) and a policy favoring full disclosure after a reasonable delay for analysis, correction, testing, and coordination. In this case the correction was recorded one day after the ZDI report, and an advisory with patch packages followed three weeks later. These notices show corrective activity, but they do not by themselves establish exploitation rates or a ranking against other firewall vendors.

References

Detect & fix
what others miss

Works with
  • GitHub
  • GitLab
  • Bitbucket
  • Azure DevOps Services
  • Gerrit
  • CVS
  • Jira
  • Linear
  • Slack
  • Security Compass
Explore integrations
Security magnifying glass visualization