A critical Puppet vulnerability is being actively exploited in attacks. The flaw, CVE-2026-115731, sits in Puppet Server’s report processing and allows a client with a valid agent certificate to execute arbitrary code on the server. The US Cybersecurity and Infrastructure Security Agency (CISA) added it to its catalog of known exploited vulnerabilities on November 17, 2026. Updates have been available since November 3. Anyone running Puppet Enterprise, Puppet Core or the community successor OpenVox should act now.
Puppet vulnerability CVE-2026-115731: What is behind it
Puppet is a widely used platform for configuration management and infrastructure as code. A central Puppet Server distributes configurations to agents on managed systems, and the agents send their results back in reports. That is exactly where the problem lies.
According to the entry in the National Vulnerability Database, the issue is deserialization of untrusted data in Puppet’s report-processing code as used by Puppet Server. A client holding any agent certificate signed by the Puppet certificate authority can submit a report containing a crafted rich-data type via a PUT request. The server then executes code with the permissions of its service user. Puppet rates the flaw 9.4 out of 10 under CVSS 4.0.
The release notes of the open source OpenVox project describe the mechanism in more detail: the server rebuilt “rich” Ruby values from the event data in a submitted report. An event carrying a crafted __ptype value therefore led to code execution on the server. Since the fix, the server reads event data with the same converter that writes it, so only plain data results.
Why autosigning becomes a risk
At first glance, an attacker needs a valid certificate. That is a weaker barrier than it sounds. Many Puppet installations automatically sign certificate requests from new agents so that new servers can be added without manual work. According to the NVD, an unauthenticated attacker can obtain such a certificate if the certificate authority autosigns requests without strong verification. OpenVox puts it more bluntly: on a server that autosigns, being able to reach the server over the network is enough.
The consequences go beyond a single server. According to vendor Perforce, as quoted by SecurityWeek, a successful attack could expose the certificate authority’s signing key. Attackers could then issue trusted certificates themselves and change the software and configuration delivered to all managed systems on their next Puppet run. A compromised Puppet Server is therefore a lever over the entire managed infrastructure.
A second critical flaw: CVE-2026-115732
Together with CVE-2026-115731, Perforce fixed a second critical vulnerability. According to the NVD, CVE-2026-115732 affects certificate-name matching in the trapperkeeper-authorization component. A client with a signed certificate whose name extends or shortens an authorized hostname is authorized as that host. This allows impersonation of other nodes. In Puppet Enterprise, the NVD says this can extend to access to certificate-authority administration and remote command execution on managed nodes. Exploitation requires a certificate with a colliding name, issued by approval or autosigning. The flaw is rated 9.0 under CVSS 4.0.
Active attacks and the CISA deadline
CISA added CVE-2026-115731 to its KEV catalog on November 17 based on evidence of active exploitation. According to the catalog entry, US federal agencies must remediate the flaw by December 1, 2026, and also follow the forensic triage requirements. The deadline does not apply directly to companies, but it is a good indicator of urgency.
According to SecurityWeek, Perforce flagged the flaw as exploited as early as November 4, one day after the updates were released. A proof-of-concept exploit targeting OpenVox was also published shortly after the patch. Who is behind the attacks and how many servers are affected is not known.
Affected versions and updates
For Perforce’s Puppet products, SecurityWeek, citing the Perforce advisory, lists the following versions as fixing both flaws:
- Puppet Enterprise and Puppet Enterprise Advanced: 2026.0.1, 2025.12.1 and 2023.8.13
- Puppet Core: 9.1.3 and 8.22.2
For OpenVox, the community fork of the no longer actively developed Puppet Open Source, the OpenVox release notes name agent versions 9.3.0 and 8.32.0 as the security releases, both dated November 3. They are part of a coordinated set with OpenVox Server 9.2.0 and 8.18.0 respectively. One detail matters: the flaw is reached through the server’s report endpoint, which runs the agent’s Ruby code inside the server. Primary servers and compilers should therefore be upgraded first. Anyone who upgrades only the openvox-agent package on a server must restart puppetserver themselves. According to OpenVox, agent-only hosts are not exposed and can be upgraded on their usual schedule.
What administrators should do now
- Patch servers first: Upgrade Puppet Server, primary servers and compilers to the versions listed and restart the service.
- Review autosigning: Turn off automatic signing or allow it only with strong verification so that only known systems receive certificates. Perforce recommends this as well.
- Audit certificates: Check the list of issued agent certificates for unknown or suspiciously named entries. Because of CVE-2026-115732, this includes names that resemble known hosts.
- Check for compromise: If the server was reachable from untrusted networks, look for unexpected reports, processes and changes to code or configuration. If the server was compromised, the certificate authority’s signing key should also be considered at risk.
- Restrict network access: Make the Puppet Server reachable only from the networks where managed systems actually reside.
Assessment
Configuration management tools are attractive targets because, by design, they distribute trust across entire server fleets. Whoever controls the Puppet Server controls what runs on the agents. The combination of easy exploitation, a public proof of concept and widespread autosigning makes CVE-2026-115731 dangerous, even though a certificate is required.
The flaw is part of a series of critical vulnerabilities in infrastructure software. Only on Tuesday, Oracle fixed Java4Shell, a flaw in OpenJDK’s Base64 decoder, and shortly before that N-able had to close 17 security flaws in its remote management platform with N-central Hotfix 3. The pattern is the same: central management tools are especially valuable targets and should be at the top of the patching list.




