They Didn't Break In Through the Front Door. They Took the Control Room.
A China-linked group chained two zero-days, BeyondTrust Remote Support and then VMware vCenter, to become a vCenter administrator and read virtual machine disks for more than two months. How it worked, why the management plane is our biggest blind spot, and what I would do on Monday.
Every week I read about a new zero-day. Most of them I note down, check whether they touch anything I care about, and move on. This week one story made me stop and read it twice.
Attackers did not phish anyone or guess a password. They used two zero-days back to back, and the second one gave them the keys to vCenter. If you have ever run a data centre, you know what that means. vCenter is the control room. Whoever holds it can see, switch off, copy or delete almost every server you own.
For years I have told teams the same thing: we protect the servers very well and forget the systems that manage the servers. This attack used exactly that gap.
From a support tool to every virtual machine, in six moves
Read it left to right. Amber boxes are places most teams have no security agent watching. MITRE ATT&CK techniques are in blue.
Here is the same story in words. The attackers first broke into a remote support appliance, the kind of tool IT uses to help colleagues. From there they moved to internal Linux appliances that had no security agent watching them, and looked around for VMware. When they found vCenter, a second zero-day let them skip the login entirely and copy the secrets that vCenter uses to sign people in.
With those secrets they simply made themselves an administrator. From then on nothing looked like an attack. They logged in to the ESXi hosts, switched on SSH, left their own keys behind and read the virtual machine disks directly, then sent the data back out the way they came in.
The tool was ready in July. The patch arrived in November.
One more detail for VMware shops: on vCenter 7.0, this fix is only available with an extended support contract.
The timeline is the part that should worry boards. The attack tool was built in July, the first break-in happened in late August, and the vCenter flaw was in use from early September. The patch arrived on 19 November. For more than two months, the only defence was noticing the attackers, because there was nothing to patch.
Where an attacker lands decides what they own
Pick a starting point. The stack shows what the attacker can reach from there, and whether anyone is usually watching.
Try a few starting points and compare. A compromised laptop is a bad day, but EDR has a decent chance of seeing it. A compromised vCenter admin is a different category: the attacker controls every host and every virtual machine, including domain controllers and backup servers if they run as VMs, and the tools inside those VMs cannot see what is happening underneath them.
What the patch fixes, and what it leaves behind
The 19 November patch closes
- ✓The vCenter authentication bypass. New attackers can no longer use CVE-2026-123890.
- ✓The door, not the room. That is all a patch was ever going to do.
If you were hit, it does not remove
- !Rogue SSO admin accounts the attackers created for themselves
- !SAML signing material they stole, which can mint new logins
- !SSH keys planted on ESXi hosts, and SSH left switched on
- !The BeyondTrust foothold they entered and left through
This is the most common mistake I expect to see after this news. Teams will apply the patch, close the ticket and move on. But if the attackers were already inside, the accounts, keys and signing secrets they created are still there after the patch. Patching stops the next intruder; only a hunt removes the current one.
What I would put on the board on Monday
The left column is for this week and you can tick items off as you go; your progress stays in your browser. The right column is the slower work that stops the next version of this attack.
Put the management plane on its own island
vCenter, ESXi management, backup consoles and remote support tools in a separate network that only jump hosts can reach.
Get eyes on the appliances
If you cannot install EDR, send their logs to your SIEM. Alert on new admin accounts, SSH being switched on and odd outbound traffic.
Treat remote support and RMM as tier-0
Same MFA, monitoring and patch urgency as your domain controllers.
Leave unsupported versions
When free fixes stop, as with vCenter 7.0, attackers know you will be slower to patch.
If someone became vCenter admin tonight, would you notice?
Five quick questions. Each “no” is a gap that would let an attacker sit in your vCenter without anyone noticing, which is exactly what happened to the victims here.
We guard the front door. They went for the control room.
This group used the front door once, through a support tool, and then went straight for the systems that run everything else. If the honest answer to “would we notice?” is “I don’t know”, that is your first project for 2027. I would love to hear how your team watches its management plane.
Sources
- Cybersecurity Dive: Chinese espionage group targets VMware vCenter
- Cybersecurity News: vCenter 0-day exploited since September
- SOC Prime: CVE-2026-123890
- Security.io: compromise hunt, not patch-only closure
- Cyber Kendra: exploited months before the patch
- ThreatDefence: patching and hunting
- SecurityElites: timeline
- HBS: This Week in Cybersecurity, 20 Nov
Facts as reported up to 21 November 2026. ATT&CK mappings are my reading of the published reports.
- VMware
- Zero-Day
- Management Plane
- Opinion