Written By: Kevin Beaumont

March 4th, VMware quietly released patches for three ESXi zero day vulnerabilities: CVE-2025–22224, CVE-2025–22225, CVE-2025–22226.

Although the advisory doesn’t explicitly say it, this is a hypervisor escape (aka a VM Escape). A threat actor with access to run code on a virtual machine can chain the three vulnerabilities to elevate access to the ESX hypervisor.

This is backed up by VMware’s official Github, which says:

Yes, this is being actively exploited in the wild.

Once you have ESX access, you can access everything on the ESX server — which includes things such as VM data, and crucially ESX config and mounted storage. Using ESX config and mounted network storage, you can traverse the VMware environment.

My pretty diagram:

For example, orgs use vMotion to allow virtual machines to automatically move across ESX hosts, to balance load and allow for maintenance without downtime (it’s how VMware security patching works). Because of this, a threat actor has direct access to storage of VMs both on and not on that host by design — they’re basically loose on the backend.

Areas of concern

ESXi is a ‘black box’ environment, where you don’t have EDR tools and such — it is locked down. As such, a hypervisor escape means a threat actor is outside of all security tooling and monitoring. They can, for example, access Active Directory Domain Controller databases without triggering any alerts anywhere in the stack, or delete data.

This is frequently seen in ransomware incidents, where people directly exploit the ESX server or vCenter server over the VMware management network using unpatched vulnerabilities. Once they reach ESX, they reach directly into storage across the whole cluster.

However, being able to reach the ESX server hypervisor directly from the Virtual Machine significantly raises the risk. For example, you don’t need to find the ESX server details, or reach a segregated network.

‘But Kevin’ you may say ‘if a threat actor gained access to a VM it’d be game over’. Well… not so much. Threat actors gain access to endpoints all the time in any large org, e.g. malware initial access on end user PCs. When you have VDIs on VMware, you have a problem. When you have shared servers on VMware, you have a problem. Compromise one of system in a company is not usually a big problem in the short term. Immediate compromise of all of them is a big problem.

Additionally, there are around 500 Managed VMware providers, who operate as effectively clouds, allowing SMBs to purchase fully managed VMs, on demand compute basically. A compromise of one customer VM would allow compromise of every customer VM in the same managed provider.

This also applies to companies who have built their own Private Clouds using VMware, and use VMware to segregate business units.

Versions impacted

The Broadcom advisory is currently incomplete for some reason. For example VMware’s Github lists versions 6.5 and 6.7 as impacted, and patches are available on VMware’s website — but VMware’s advisory on the Broadcom site doesn’t list them as impacted as of writing. Basically, every release of ESX is impacted.

I understand 5.5 is also impacted, however it is out of support so no patch is available.

Continue reading article here!