# Puppet Server flaw CVE-2026-115731 is now exploited, says CISA: an agent certificate can mean code execution on your config server

> CISA added the Puppet report-deserialization bug to its Known Exploited Vulnerabilities list on November 17, two weeks after Perforce and OpenVox shipped fixes. Any machine with a signed agent certificate, or anyone at all if autosigning is loose, can run code on the server that holds your Puppet CA key.

Published Nov 17, 2026 by TechMeasured Staff · Security · Canonical: https://techmeasured.com/security/puppet-server-cve-2026-115731-report-rce-cisa-kev-exploited-upgrade-autosign-what-to-do

The US Cybersecurity and Infrastructure Security Agency (CISA) added **CVE-2026-115731**, a code-execution flaw in Puppet Server's report handling, to its Known Exploited Vulnerabilities (KEV) catalog on November 17, citing evidence of active exploitation, according to WindowsForum and nFlo's KEV tracker. Fixes for Puppet Enterprise, Puppet Core and the community fork OpenVox have been out since November 3. nFlo lists **December 1** as the remediation deadline for US federal agencies.

Why it matters: the Puppet server is the machine that tells the rest of your fleet what to run. SecurityOnline notes that code execution as the Puppet service user gives access to the CA signing key. That key is what the rest of your fleet trusts. A public proof of concept has been available since disclosure.

## What the bug is

| Detail | What's reported |
|---|---|
| Flaw | Deserialization of untrusted data in report processing (CWE-502) |
| Trigger | A crafted "rich data" type in a report sent by `PUT /puppet/v3/report` |
| Who can exploit | Any client with a CA-signed agent certificate; potentially unauthenticated attackers if the CA autosigns requests without strong checks (Red Hat) |
| Result | Code execution as the Puppet Server service user |
| Severity | CVSS 4.0: 9.4 (Perforce/CNA); CVSS 3.1: 9.9 (Red Hat) |
| Disclosed and patched | November 3, 2026 |
| Added to CISA KEV | November 17, 2026 |
| Proof of concept | Public, in the OpenVox advisory GHSA-ffr6-rwpm-p5wq (SecurityOnline) |

SecurityOnline says the server trusts an attacker-supplied type name that reaches a Ruby `class_eval` call. Red Hat notes that inside Red Hat Satellite environments the affected service processes incoming client reports over the network by default. Red Hat's page doesn't yet list a fixed Satellite package.

CISA hasn't said who is exploiting the flaw or how, and the reports we read give no attacker details. Coverage around disclosure was inconsistent. SecurityOnline's article, published before the KEV listing, said no in-the-wild exploitation of the RCE had been confirmed. It labelled a companion bug as exploited, based on an unverified Reddit thread.

## The companion bug

The same releases fix **CVE-2026-115732** (CVSS 9.0), a certificate-name authorization bypass in the trapperkeeper-authorization library. It treats hostname rules like wildcards, so one node can impersonate another. In Puppet Enterprise this can lead to remote command execution on managed nodes, per SecurityOnline. It is not on the KEV list, according to WindowsForum, but the same upgrade fixes it.

## Fixed versions

| Product | Fixed in |
|---|---|
| Puppet Enterprise 2026.0 | 2026.0.1 (Puppet Core 9.1.3) |
| Puppet Enterprise 2025.12 | 2025.12.1 (Puppet Core 8.22.2) |
| Puppet Enterprise 2023.8 | 2023.8.13 (Puppet Core 8.22.2) |
| OpenVox Server | 9.2.0 or 8.18.0 |
| OpenVox Agent | 9.3.0 or 8.32.0 |
| OpenVoxDB (incl. container image) | 9.2.0 or 8.18.0 |
| Open Source Puppet | No further updates, per SecurityOnline |

Sources: WindowsForum (Perforce lines) and SecurityOnline (OpenVox). If you still run the abandoned open-source Puppet packages, the realistic fixes are moving to OpenVox or to Puppet Core/Enterprise.

## What to do

1. **Upgrade the server now** to the fixed release for your line, then restart Puppet Server. Patching is the only full fix.
2. **Turn off or tighten autosigning.** Loose autosigning turns this from "a compromised agent can attack the server" into "anyone who can reach the server can," per Red Hat. Red Hat's interim advice is to never autosign without strong verification.
3. **Restrict network access** to Puppet Server so only managed nodes can reach it. Limiting `PUT /puppet/v3/report` helps, but WindowsForum stresses this reduces exposure rather than fixing the flaw.
4. **Hunt for signs of compromise,** especially if the server was reachable from untrusted networks before November 3:
   - unusual commands or child processes run by the `puppetserver` user (EDR or auditd logs)
   - recently signed certificates and agent registrations you don't recognise
5. **If you find compromise, treat the CA as burned.** Rotate the Puppet CA and any secrets the server handled, and preserve logs before rebuilding.
6. **After upgrading, review custom `auth.conf` rules.** SecurityOnline warns that some rules may have depended on the buggy wildcard behaviour fixed by CVE-2026-115732.

Federal agencies have until December 1 to remediate under CISA's directive. For everyone else, a two-week-old patch with a public PoC and confirmed exploitation should go to the top of the queue.


## Sources

1. [Red Hat Customer Portal — CVE-2026-115731](https://access.redhat.com/security/cve/CVE-2026-115731)
2. [WindowsForum — CVE-2026-115731: CISA Warns of Active Exploitation of Puppet Server Code Execution Flaw](https://windowsforum.com/news/cve-2026-115731-cisa-warns-of-active-exploitation-of-puppet-server-code-execution-flaw.452778/)
3. [SecurityOnline — Puppet Server RCE CVE-2026-115731 PoC](https://securityonline.info/puppet-server-rce-cve-2026-115731-poc/)
4. [nFlo — CVE-2026-115731 in Perforce Puppet Enterprise](https://nflo.tech/knowledge-base/2026-11-17-cve-2026-115731-en/)
