Puppet Server Report Flaw Exploited (CVE-2026-115731): Patch Now

Rescana advisory card: Puppet Server report deserialization remote code execution (CVE-2026-115731), actively exploited, listed in CISA KEV. Patch now.

Status: CVE-2026-115731 is in CISA's Known Exploited Vulnerabilities (KEV) catalog, added November 17, 2026, with a federal due date of 1 December 2026. It lets anyone holding a certificate signed by your Puppet certificate authority (CA) run code on Puppet Server by sending a crafted agent report. Perforce fixed it on 3 November 2026 in Puppet Enterprise 2026.0.1, 2025.12.1 and 2023.8.13 and Puppet Core 9.1.3 and 8.22.2. The Centre for Cybersecurity Belgium (CCB) says the flaw is "actively exploited in the wild" and that a public proof of concept is available. The main limit on attackers is the certificate: without a CA-signed agent certificate there is no attack, unless your CA autosigns requests without strong checks. A companion flaw fixed in the same releases, CVE-2026-115732, is not in KEV. We last checked at 13:28 EST on 19 November 2026.

Executive Summary

Puppet Server is the control point of a Puppet estate: it compiles and serves the configuration every managed server and workstation applies, and by default it also runs the certificate authority that decides which machines are trusted. Agents send a report back to the server after each run. CVE-2026-115731 is a flaw in how the server processes those reports.

Perforce's description: a client that holds "any CA-signed agent certificate" can "execute arbitrary code on the Puppet Server as its service user by submitting a report containing a crafted rich-data type via PUT request". Perforce adds that "Where the certificate authority autosigns requests without strong verification, an unauthenticated network attacker can obtain such a certificate."

Perforce's knowledge base article spells out why this matters: a successful attack "could expose the server's certificate-signing key, allow an attacker to issue trusted certificates, and enable changes to software and configuration delivered to managed systems on their next Puppet run." In other words, the Puppet Server is a path to every machine it manages.

Three things decide whether a Puppet Server is exposed:

  1. An affected build. Puppet Enterprise 2026.0.0, 2025.0.0 up to 2025.12.0, or 2023.8.0 up to 2023.8.12; Puppet Core 9.0.0 up to 9.1.2 or 8.x up to 8.22.1.
  2. A CA-signed agent certificate. Any agent certificate works. That means any managed machine, or anyone who has stolen one machine's certificate and private key, can attempt it. An outsider can only get a certificate if your CA autosigns requests without strong verification.
  3. Network reach to Puppet Server. The attacker must be able to connect to the server's agent port (TCP 8140 by default).

How serious is it in practice? (Rescana guidance) The certificate requirement is a real de-escalating fact. Estates that sign certificates by hand, or with a policy script that checks a shared secret or cloud identity, are only exposed to attackers who already control a managed machine. Two things push it back up. First, an attacker who compromises one managed machine can use this flaw to take control of the whole estate. Second, autosigning can be used to bootstrap cloud and container fleets, and Perforce's own guidance says basic allowlist autosigning is insecure because "any host can request any certname". Treat any server with naive or basic autosigning as exposed to unauthenticated attackers on the network.

Exploitation status. CISA added CVE-2026-115731 to KEV on 17 November 2026 and lists ransomware use as "Unknown". CISA's SSVC entry on the CVE record rates exploitation "active", automatable "no" and technical impact "total". CCB's 18 November warning says the flaw is "actively exploited in the wild" and that a public proof of concept is available. SecurityWeek reports that proof-of-concept code targeting OpenVox was published shortly after the 3 November fix; the published demonstration only creates a file on the server, and no source we found ties it to the attacks CISA recorded. We have not opened or reproduced it. Perforce's advisory does not describe exploitation, and no source we checked publishes indicators of compromise.

Technical Information

The flaw (CVE-2026-115731). CWE-502, deserialization of untrusted data, plus CWE-94, code injection. CCB's technical summary: after each run, agents send a JSON report to the server, and the server converts the report's event data with Puppet's rich-data deserializer (FromDataConverter), which "rebuilds data types from an attacker-controlled __ptype field. A crafted type definition reaches a code path that evaluates attacker-supplied Ruby code". The attacker submits the report to the report endpoint with a PUT request, and code runs as the Puppet Server service user.

Severity. Perforce rates it 9.4 Critical under CVSS 4.0, with low privileges required: the low privilege is the agent certificate. NVD's own analysis rates it 9.9 Critical under CVSS 3.1 with a changed scope, reflecting that compromise of the server reaches the machines it manages.

The companion flaw (CVE-2026-115732). Certificate name authorization bypass, CWE-863 and CWE-697, rated 9.0 Critical under CVSS 4.0 by Perforce. Perforce says the certificate-name matching in trapperkeeper-authorization lets "a client that holds a CA-signed certificate whose name extends or shortens an authorized hostname" be authorized as that host. Perforce says that enables impersonation of other nodes and, in Puppet Enterprise, "access to certificate-authority administration and remote command execution on managed nodes". Exploitation needs a certificate with a colliding name, which the CA issues by approval or autosign. It is not in KEV, and CISA's SSVC entry rates its exploitation "none". The same upgrade fixes both.

Timeline (EST).

  • 3 November 2026: Perforce and OpenVox publish fixed releases and advisories; the CVE records for CVE-2026-115732 and CVE-2026-115731 follow at 15:38 and 15:40 EST.
  • 3 November 2026: OpenVox's advisory includes technical detail sufficient to reproduce the flaw; proof-of-concept code is public from about this date.
  • 17 November 2026: CISA adds CVE-2026-115731 to KEV, due 1 December.
  • 18 November 2026: CCB publishes its warning that the flaw is actively exploited.

Affected Product Versions

Product Affected Fixed in
Puppet Enterprise and PE Advanced 2026.0 2026.0.0 2026.0.1 (Puppet Core 9.1.3)
Puppet Enterprise and PE Advanced 2025.x 2025.0.0 up to 2025.12.0 2025.12.1 (Puppet Core 8.22.2)
Puppet Enterprise and PE Advanced 2023.8 LTS 2023.8.0 up to 2023.8.12 2023.8.13 (Puppet Core 8.22.2)
Puppet Core 9 9.0.0 up to 9.1.2 9.1.3
Puppet Core 8 8.x up to 8.22.1 8.22.2

Sources: Perforce's CVE page and knowledge base article, the CVE record and the 8.22.2, 9.1.3, 2023.8.13 and 2025.12.1 release notes. The CVE record adds that you must "restart puppet server so the fixed code is loaded."

Core 8.22.1 or 8.22.2? The fixed Puppet Core 8 release is 8.22.2. Perforce's CVE page, its knowledge base article ("Puppet Core: 8.22.1 and older" affected; fixed in "9.1.3 or 8.22.2"), the CVE record, NVD and CCB all agree, and the Puppet Core 8.22.2 release notes list both CVEs as fixed. The "8.22.1" figure comes from a heading on puppet.com's What's New page, which links to the 8.22.2 release notes. Puppet Core 8.22.1 is affected.

OpenVox. OpenVox, the community fork of open-source Puppet, shipped fixes the same day. Its release notes say the same report issue "in Puppet Enterprise is CVE-2026-115731", and the OpenVox project lists the fixed versions as openvox-server 9.2.0 and 8.18.0, openvox-agent 9.3.0 and 8.32.0, and openvoxdb 9.2.0 and 8.18.0, "or greater". OpenVoxDB also carries the CVE-2026-115732 fix, so upgrade it too (OpenVoxDB release notes; OpenVox release notes; OpenVox announcement on oss-security). If you upgrade only the agent package on a server, restart puppetserver yourself. Alain of Neodyme found and reported both flaws, according to Neodyme's write-up.

Open-source Puppet and older releases. OpenVox's oss-security announcement (OpenVox announcement on oss-security, 3 November 2026) says the two flaws "affect all known OpenVox and Puppet versions and implementations", and that "Puppet will not be receiving security updates" for open-source Puppet. Perforce's fixes cover only Puppet Enterprise and Puppet Core, and the CVE record's affected ranges start at Puppet Enterprise 2023.8.0. Neodyme, which found the flaw, says the exploit also works against open-source Puppet 6.24.0 from 2021. Treat open-source Puppet servers, distribution-packaged Puppet and end-of-life Puppet Enterprise releases as affected with no Perforce fix, and move them to a fixed Puppet Core, Puppet Enterprise or OpenVox release (Rescana guidance).

How to Find Your Exposure

The steps below are Rescana guidance unless a source is named.

1. Inventory every Puppet Server and CA. Include compilers and replica servers in Puppet Enterprise, and any separate Puppet Core servers run by teams outside central IT. Record the version on every server, not just the primary. In Puppet Enterprise, cat /opt/puppetlabs/server/pe_version gives the PE release. On every Puppet Server, cat /opt/puppetlabs/puppet/VERSION, rpm -q puppet-agent (Red Hat family) or dpkg-query -W puppet-agent (Debian family) gives the bundled agent version; 8.22.2, 9.1.3 or later is fixed. puppetserver --version alone does not show the fix, because it ships in the agent package (Rescana reading of the CVE record). Packages from a Linux distribution's own repositories, rather than Perforce or OpenVox repositories, are open-source builds with no Perforce fix. For OpenVox, the release notes say "Agent-only hosts are not exposed by the issue and can upgrade on their usual schedule", so start with servers and compilers.

2. Check autosigning (Perforce). On the CA primary, run puppet config print autosign --section server. Perforce's article explains the result: true means naive autosigning, where "every CSR is signed automatically"; false means autosigning is disabled; a file path means basic allowlist autosigning, or policy-based autosigning if the file is executable. In PE, an empty or missing autosign.conf means nothing is autosigned. Review the entries in /etc/puppetlabs/puppet/autosign.conf if it exists.

3. Rank servers. First: affected servers with naive or basic autosigning reachable from networks you do not fully control. Next: affected servers with policy-based autosigning; check what the policy script validates. Then: servers with manual signing, which are exposed to anyone who controls a managed machine.

4. Check reachability. List which networks can reach Puppet Server's agent port (TCP 8140 by default), including cloud VPCs, partner links and VPN pools.

5. Confirm after upgrading. Check the new version and confirm Puppet Server was restarted, as the CVE record requires.

Workaround and Mitigation

Patch. Upgrade to PE 2026.0.1, 2025.12.1 or 2023.8.13, or Puppet Core 9.1.3 or 8.22.2, then restart Puppet Server. Perforce calls its mitigations "stop-gap measures" that "do not replace upgrading". OpenVox advises upgrading "primary servers and compilers first"; in Puppet Enterprise, upgrade the primary, every compiler and any replica (Rescana guidance).

Perforce's temporary mitigations for CVE-2026-115731:

  • "Turn off autosigning or enforce strong validation in autosign scripts so only trusted nodes receive certificates." Perforce's article gives the steps: set autosign = false in the [server] section of puppet.conf and reload Puppet Server (systemctl reload pe-puppetserver in PE).
  • "Apply strong access controls to Puppet agent hosts so unauthorized users cannot access signed certificates or private keys."

Limits of these mitigations (Rescana guidance): turning off autosigning stops outsiders getting new certificates, but it does not stop anyone who already holds a valid agent certificate, including certificates autosigned before the change. Review autosigned certificates and revoke any you cannot attribute. Turning off autosigning also breaks automated enrolment for new cloud or container nodes until a policy script is in place, so plan the bootstrap change with the platform team.

For CVE-2026-115732, Perforce adds: apply strict controls so certificates cannot be issued with names that extend approved names, and review your trapperkeeper-authorization (tk-auth) rules. Perforce warns that after upgrading, tk-auth wildcard matching is stricter: a rule for a wildcard domain no longer matches the base domain itself, so review and test custom wildcard rules.

Additional steps (Rescana guidance): restrict network access to Puppet Server's agent port (TCP 8140 by default) to the networks where managed machines live; and if your fleet bootstrap depends on autosigning, move it to a policy script that verifies something an attacker cannot forge, such as a cloud instance identity document or a one-time token.

Indicators of Compromise

Perforce, CCB and CISA have published no indicators of compromise. CCB notes that patching "does not remediate historic compromise". The checks below follow from the impact Perforce describes (code execution as the service user, exposure of the CA signing key, rogue certificates and changed configuration) and are Rescana guidance, not vendor IoCs:

  • Certificates. Review the CA's signed certificate inventory for names you do not recognise, especially any signed by autosign, and for names that extend or shorten real hostnames (the CVE-2026-115732 pattern).
  • Reports (hunt window: 3 November 2026, when the fix and OpenVox's technical advisory went public, to today). On the primary and every compiler, search /var/log/puppetlabs/puppetserver/puppetserver-access.log (the default location in Perforce's log reference), including rotated copies, for PUT /puppet/v3/report/ requests from certificate names that are not real managed machines, from unexpected source addresses, or from certificates autosigned recently. Check /var/log/puppetlabs/puppetserver/puppetserver.log for report-processing errors in the same window. If you keep reports on disk (puppet config print reportdir --section server) or in PuppetDB, search them for event data carrying a __ptype value with an inline type definition that begins with the word Object and treat any hit as suspicious. A failed attempt may leave only the access log line, and logs past your retention period mean unknown, not clean.
  • Server host. Look for unexpected processes started by the Puppet Server service account (pe-puppet in Puppet Enterprise, puppet in Puppet Core and OpenVox) and for unexplained files in its directories.
  • Delivered configuration. Review recent changes to code, modules, Hiera data and classification for anything not in your change control.
  • If you find signs of compromise. Treat the CA signing key as exposed. Rebuilding the CA and reissuing agent certificates is disruptive, so plan it with Perforce support, and rotate any secrets Puppet distributes to managed machines. A stolen CA signing key lets an attacker mint certificates offline that never appear in the CA's signed-certificate inventory, so a clean inventory does not rule out compromise. In Puppet Enterprise, the reporter showed that a stolen CA key can be used to reach the orchestrator broker (TCP 8142) and run commands as root on managed nodes, so review orchestrator job and task history first for runs you did not start.

Third-Party Risk (TPRM) Angle

Puppet can run inside suppliers you depend on, such as managed service providers, hosting and cloud operators, SaaS companies running their own fleets, and outsourced IT. A compromised Puppet Server lets an attacker push configuration to every machine it manages, so a supplier's Puppet estate is effectively a path into the services and data they run for you (Rescana guidance).

The risk concentrates in three supplier types:

  • MSPs and outsourced IT providers that run Puppet to manage customer servers, where one Puppet Server may touch several customers.
  • Hosting, cloud and SaaS providers that bootstrap large fleets, where autosigning may be in use.
  • Suppliers running open-source Puppet or end-of-life releases, which get no Perforce fix, or Puppet Core outside a support contract, who may not track Perforce advisories.

Send these three questions to suppliers that run Puppet in the environments that host your data or services (Rescana guidance):

  1. Which Puppet Enterprise or Puppet Core version runs on your Puppet Servers, and when did you upgrade to PE 2026.0.1, 2025.12.1 or 2023.8.13, or Core 9.1.3 or 8.22.2 (for OpenVox, openvox-server and openvoxdb 9.2.0 or 8.18.0)?
  2. Does your Puppet CA autosign certificate requests, and if so, what does the autosign policy verify before it signs?
  3. Have you reviewed your signed certificates, report logs and recent configuration changes for signs of misuse since 3 November, and what did you find?

Ask for evidence, not reassurance: the version on each Puppet Server, the output of the autosign check, and the date of the certificate review. Record answers against each supplier, and follow up with any supplier still on an affected build or relying on basic allowlist autosigning.

What to do internally: find every Puppet Server, upgrade and restart it, turn off weak autosigning until then, and check signed certificates and recent configuration changes. We will update this advisory if Perforce or CISA publishes indicators, if CVE-2026-115732 is added to KEV, or if ransomware use is reported.

Need help working out which suppliers run Puppet servers exposed to CVE-2026-115731? Our team can help you prioritize them and track their answers. Talk to an expert

Sources

Talk to an expert

Talk to Rescana about this advisory — or about how we help TPRM teams keep reviews moving.