Blog Article

Certighost (CVE-2026-54121): Why ADCS Hardening Isn't Enough

November 5, 2026
7 min read
Expert Content

Published on

November 5, 2026

In July 2026, Microsoft patched a flaw in Active Directory Certificate Services (AD CS) that deserves more than a line in a Patch Tuesday report. The researchers who found it named it Certighost. Tracked as CVE-2026-54121, it let an ordinary domain user obtain a certificate that identifies them as a Domain Controller.

Most AD CS security stories start with a badly configured certificate template. This one does not. According to Dataminr, the attack relied on the default Machine template and default domain settings that are common in unmodified deployments. That is why it is worth a closer look for anyone who runs an enterprise PKI.

The lesson is not that AD CS is broken. Microsoft fixed the issue after responsible disclosure, and the patch works. The lesson is about what hardening can and cannot do, and why the shape of your PKI matters as much as its settings.

What Certighost demonstrates

In a Windows domain, a certificate works like a login badge. Through a mechanism called PKINIT, a user can sign in to Kerberos with a certificate instead of a password. AD CS is the badge printer. If it prints a badge with someone else's name on it, the holder walks in as that person.

Certighost tricks the printer. When a Certificate Authority (CA) processes an enrollment request, it looks up the requester's identity in the directory. According to Help Net Security, in some cross-domain scenarios the CA runs a second lookup, which the researchers call a "chase." Two values in the request, cdc (Client DC) and rmd (Remote Domain), tell the CA which host to contact and which account to look up.

The problem is that the requester controls both values. As Help Net Security explains, the attacker points the CA at a host they control and serves it the identity of a real Domain Controller. To pass the CA's checks, they use a machine account. Any domain user can create one under the default ms-DS-MachineAccountQuota setting.

The CA then signs a certificate that names the Domain Controller. With it, the attacker authenticates as that DC and runs DCSync to extract the krbtgt secret. That is full control of the domain. And because a forest, not a domain, is the real security boundary in Active Directory, a compromised child domain is a short step from the forest root.

The timeline moved fast. According to Dataminr, the flaw was reported to Microsoft on May 14, 2026, and patched on July 14, 2026. A working public exploit followed on July 24, 2026. The vulnerability carries a CVSS score of 8.8.

Two kinds of flaw, two kinds of fix

Since SpecterOps published its Certified Pre-Owned research in 2021, security researchers have catalogued a long family of AD CS escalation paths. They are now known as ESC1 to ESC17. Open-source tools such as Certipy and Certify are built to find and abuse them.

Nearly all of them share one trait: they are misconfigurations. A template lets users supply their own subject name. A CA flag accepts alternative names sent as request attributes. An AD object carries loose permissions. A domain controller accepts weak certificate mappings. Each one is a setting, and a fully hardened AD CS is not exposed to any of them.

Certighost belongs to a different category. No template was misconfigured and no permission was too broad. The flaw sat in the CA's own enrollment logic, in the code that decides who a request really comes from.

Take control of your PKI infrastructure

See how Evertrust simplifies certificate lifecycle management.

Get Started

The researchers did describe a temporary workaround: a CA policy flag that switches the chase fallback off. But before July 2026, no hardening guide told anyone to touch it, because nobody knew it mattered. You cannot configure your way out of a flaw you do not know exists. You patch it, or you remove the component that contains it.

Why conventional hardening has limits

Hardening is essential, and nothing here argues against it. Auditing templates, scoping enrollment rights, enforcing strong certificate mapping and locking down CA permissions closes most known AD CS attack paths. Every organisation running AD CS should do it.

But hardening has a built-in limit. It protects you against what is already known. A checklist is a record of past mistakes, and Certighost was not a mistake anyone made. It was an assumption inside the software that turned out to be wrong.

There is also the question of scale. A typical AD CS estate spreads its security across certificate templates, CA flags, AD object permissions, registry keys on domain controllers and the CA server itself. Each of these drifts on its own schedule. Every new template is one more chance to get a setting wrong. Keeping it all clean is possible, but it is a permanent effort.

So the question for a security team is not only "are we hardened?" It is also "what happens on the day the CA itself has a flaw?"

Patching first, architecture next

The first step is simple: apply the July 2026 security update to every AD CS server. According to Help Net Security, the fix makes the CA validate the chase target before the lookup goes ahead. Dataminr also recommends lowering the machine account quota and reviewing CA logs for certificates issued to Domain Controller identities through unusual accounts.

Patching closes this flaw. It does not change the exposure model. The next code flaw will follow the same path: disclosure, patch, then a race to deploy. Architecture is what decides how much damage a flaw can do before the patch lands.

Put one governed door in front of the CA

A Registration Authority (RA) sits between requesters and the CA. A strict RA does not forward a request as it arrives. It rebuilds every certificate from an enforced policy. Used as an RA in front of AD CS, Evertrust CLM works this way. It enrolls on behalf of users, under one fixed template, and discards the requested subject, names and extensions before rebuilding the request from approved values.

Want to master certificate management?

Browse our resources on PKI best practices.

Education Center

For Certighost, this matters. A low-privileged user cannot steer the CA's identity lookup through the RA, because they never talk to the CA directly. The same design closes several classic ESC paths by construction rather than by configuration: no template shopping, no injected key usages, no alternative names smuggled in through request attributes.

There is one honest condition. An RA only protects the path that goes through it. If attackers can still reach the CA directly, the old exposure remains and the CA still needs its patch. The gain is real when the RA becomes the only door.

Remove the component, remove the risk

The durable step is to replace what stands behind that door. With Evertrust CLM in place, moving issuance to Evertrust PKI, a CA with no Active Directory integration, takes AD CS out of the picture. There are no AD-published templates, no CA policy flags, no web or RPC relay endpoints, and no directory lookup logic for a Certighost-class flaw to abuse.

This is not a magic box either. If Evertrust PKI is trusted for smart card logon, its identity fields must stay locked: keep the SID non-editable and write template allow-lists tightly. The difference is scale. What remains is a short list of explicit, reviewable decisions, not a sprawling surface spread across templates, flags, permissions and server code.

Contain, then eliminate

Certighost is a reminder that a fully hardened CA is still software, and software can fail. Hardening remains the baseline. Patching remains non-negotiable. Neither changes the fact that a single flaw in the CA can hand over the forest.

Architecture does. A strict Registration Authority turns a broad attack surface into one governed, audited door. Removing the AD CS dependency behind it turns containment into elimination.

Want to know where your AD CS environment stands? Talk to our team about how Evertrust CLM and Evertrust PKI fit your architecture.

Sources

Was this helpful?
Back to blog

Table of Contents

Stay Updated

Get the latest PKI insights delivered to your inbox.

Subscribe

By subscribing you accept to receive our communications. You can unsubscribe at any moment.

Related Articles

Evertrust PQC

Are European enterprises ready for Post-Quantum Cryptography (PQC) migration? The gaps and the path forward

September 10, 2025
1 min

Explore why PQC adoption lags in Europe, the real blockers, and how to achieve quantum-safe security.

Read more
Evertrust PQC

NIST Releases New Post-Quantum Cryptography Standards

September 10, 2025
1 min

Discover NIST’s new Post-Quantum Cryptography standards (FIPS 203, 204, 205) and how Evertrust is preparing to integrate them for enhanced cybersecurity.

Read more
Evertrust ACME

ACME Clients on Linux

February 12, 2024
1 min

The ACME protocol is a network protocol designed to automate the process of domain validation, deliverance and renewal of X.509 certificates. The process is set up between an ACME server and an ACME client.

Read more
Get started

Ready to take back control over your certificates?

Talk to our experts and discover how Evertrust can help you implement best practices in PKI and certificate lifecycle management.