On 17 November 2026, CISA added CVE-2026-115731 to its Known Exploited Vulnerabilities catalog (catalog version 2026.11.17). The flaw is a deserialization of untrusted data issue in Perforce Puppet Enterprise's report-processing code: anyone holding a CA-signed agent certificate can submit a report containing a crafted rich-data type over a PUT request and execute arbitrary code on the Puppet Server as the service user. If your certificate authority auto-signs requests without strong verification, an unauthenticated network attacker can obtain that certificate and do the same. The fix is the vendor's mitigations, applied to every Puppet Server you run, ahead of the 1 December 2026 due date.
What CVE-2026-115731 actually is
It is a deserialization of untrusted data vulnerability in Puppet's report-processing code, catalogued by CISA as the "Perforce Puppet Enterprise Deserialization of Untrusted Data Vulnerability."
Puppet agents send reports back to the Puppet Server as a normal part of operation. That code path parses data the server did not create, which is the classic shape of a deserialization bug: untrusted bytes go in, and an object graph comes out with more authority than the sender should have had. CISA's description says the crafted rich-data type arrives through a PUT request and results in code execution as the service user. Because report submission is routine traffic, the request will not look as obviously wrong as a stray POST to an admin endpoint would.
CISA lists known ransomware campaign use as unknown. That means no confirmed link, not a clean bill of health.
Am I affected? Your certificate authority decides
You are affected if you run Puppet Enterprise and your CA will sign certificate requests from sources you do not strongly verify.
The harder version of this attack needs a certificate, and an already-enrolled agent has one — so a compromised node or a careless insider is enough. The easier version needs your CA to auto-sign without strong verification, which turns an authenticated bug into an unauthenticated one for anyone who can reach the signing endpoint. CISA puts the internet-exposure judgement on the asset owner, so check whether your Puppet Server is reachable from outside. Internal-only reduces exposure; it does not remove it, since an agent certificate is a low bar inside most networks. Until you have confirmed the certificate path is locked down, treat every Puppet Server as in scope.
What to do in the next 24 hours
Apply the vendor's mitigations to every Puppet Server you run, then verify what is actually answering traffic — not what you think you deployed.
- Pull the current Perforce advisory for Puppet Enterprise and follow its mitigation steps. CISA's entry points at vendor instructions and does not itself name a fixed version, so the advisory is the source of truth here.
- Inventory every Puppet Server, including staging, DR, and the VM a contractor built and nobody documented. These are usually where the gap is.
- Confirm the version on the host or pod sitting behind the load balancer, not on the golden image or the repo. In containers, compare the running image digest against the tag you believe you shipped.
- If your CA auto-signs, disable that or add strong verification first. It is the single setting that decides whether this is an authenticated or unauthenticated bug for you.
- If no mitigation is available to you, BOD 26-04's guidance for cloud services applies, and CISA names discontinuing use of the product as the alternative.
The 2026-12-01 due date is binding on federal agencies; there is no reason to give yourself a longer one. Teams running Puppet and other config-management estates usually find the CA setting is the part nobody owns.
What to watch in the logs afterwards
Watch the report-submission path and the certificate issuance path, because the exploit needs both halves.
On the Puppet Server, examine PUT requests carrying report payloads — particularly from certificate names that do not map to a node in your inventory, or from agents that do not normally report. On the CA side, review every certificate signed during your exposure window, auto-signed ones first, and reconcile the names against your node list. Then look at behaviour attributable to the service user: processes the Puppet Server account should not be spawning, outbound connections it should not be making, files written outside its expected directories. CISA also references its Forensics Triage Requirements, which is the checklist to work through if you find something concrete rather than merely suspicious.
What AI-assisted review changes about bugs like this
Deserialization flaws are a data-path tracing problem, and that is where AI-assisted code review genuinely helps — it does not help with the design decision that caused the problem.
An engineer working with an assistant can follow untrusted input from entry point to type resolver far faster than reading the same files line by line, and can ask directly for every place a report payload reaches object construction. That compresses the time between an advisory landing and you understanding whether your own custom report handlers, plugins or integrations are exposed. What AI does not fix is a report pipeline that trusts whatever an agent sends, which is a design choice and still a human one to make. If your team maintains its own handlers, this is a good week to have a model and a reviewer read them together. The advisory does not say how this particular bug was found, and no claim is made here.
Where to start
Start with the certificate authority, not the patch. Turn off auto-signing or add strong verification, inventory the Puppet Servers, then apply the mitigations and confirm the version on the box answering traffic. If you want a second pair of eyes across the Puppet estate, the CA configuration, and the internal tooling around them, start here — Maysan Labs does this kind of infrastructure and platform review, and for the wider picture see our cloud and infrastructure work.
Buying software or a website?
Twelve questions to ask any vendor before you sign: price, your data, fit for an Indian business and support.
Get the free checklistWritten by Maysan Engineering Team at Maysan Labs, a software studio in Gurgaon, India. Follow us on LinkedIn