A VMware estate wiped with no exploit, just logins
Kaspersky GERT documented an attack that destroyed four ESXi hosts, both vCenter appliances and the backups in ten days, using only stolen admin credentials.
By Yaali. November 16, 2026, 6 min read, Ransomware, Resilience.
Over roughly ten days an attacker took apart an organisation's entire VMware virtualisation estate without exploiting a single software flaw. Kaspersky's Global Emergency Response Team (GERT), the unit it sends into live incidents, published its account on 13 November 2026, with the victim left unnamed. By the end, four VMware ESXi hosts and every virtual machine (VM) running on them were gone, both vCenter appliances that managed those hosts were encrypted, and the connected backup infrastructure had already been destroyed. A VM is a whole computer running as software on a physical server; ESXi is VMware's layer that hosts those VMs, and vCenter is the console that manages many ESXi hosts at once.
Anyone running VMware should take this seriously, because the whole intrusion ran on stolen identities. Nothing in it depended on an unpatched product. The attacker used valid administrator credentials, stolen from systems outside the virtualisation layer, then worked through VMware's own management features. Keeping your hypervisors up to date would not have stopped it. The controls that would have are separated admin identities and a backup copy those credentials could not reach.

How it works
A hypervisor is the software that runs VMs directly on server hardware. ESXi is VMware's bare-metal hypervisor, meaning it installs on the physical machine itself: one ESXi host can run dozens of VMs, each with its own operating system and virtual disks. vCenter Server is the central manager, and it holds administrative control over every ESXi host joined to it. That design is also the weak point. Whoever holds root on vCenter controls every host beneath it, and through those hosts, every VM.
Host Profiles is the feature the attacker turned against the estate. It is a vCenter tool for configuration management: an administrator captures a reference configuration and pushes it out to keep many hosts consistent. The same mechanism can push a setting that enables SSH (Secure Shell, the remote command-line service) on the ESXi hosts. Once that is on, the hosts trust connections coming from vCenter, so vCenter becomes a path onto each host without the attacker ever needing the hosts' own passwords.
Working at the hypervisor layer is why so little effort destroyed so much. Deleting or overwriting a VM's disk files on an ESXi host wipes that whole machine, and a handful of hosts carry an organisation's servers. Destroying the backup infrastructure first removed the only route back. By the time the production VMs went down, there was nothing left to restore them from, and all of it happened with legitimate admin access rather than any exploit.
What the attacker did
The chain started outside the virtualisation layer. Credentials harvested from ordinary systems inside the network were used to reach a vCenter appliance, where the attacker logged in as root over SSH. Within about 43 seconds of that login, a script deployed MeshAgent, a remote management agent, for persistence. Roughly four hours later the attacker returned through MeshAgent and installed Ligolo-ng, a tunnelling tool that opens a reverse connection out of the network, running it as a disguised service named to look like a normal VMware component.
From vCenter the attacker used Host Profiles to switch on SSH across the ESXi hosts, then connected to each host from the trusted vCenter address. The ransomware payload was delivered over SFTP, made executable, and run first with a self-test switch before the real detonation about three hours later. On one host, every VM stopped within 23 seconds, and the first few kilobytes of each virtual disk were overwritten, enough to break the partition tables.
The sequence is the part worth copying into your own threat model. The backup environment was destroyed first, not last. Days before the detonation, the attacker disabled antivirus (AV) on the backup servers, then stopped the backup services and encrypted the recovery repositories. Only after the recovery path was gone did the production hosts and the vCenter appliances go down.
The group behind it is not fully confirmed. The ransom notes named BlackNevas, a family first seen in November 2024 that SentinelOne and others describe as derived from Trigona ransomware, with builds that target ESXi. Kaspersky could not confirm the family through binary analysis, because the ESXi payload itself was never recovered, so there is no hash or reverse-engineered match. The link rests on the style of the command-line switches used at detonation, which resemble that family's. Treat the BlackNevas label as a likely attribution, not a proven one. The mechanics of the attack stand on their own regardless of who ran it.
What to do

Work through these in order.
- Treat vCenter and ESXi management as Tier 0. Tier 0 is the set of systems that control the rest of the estate, held to the strictest access rules, the way domain controllers are. Give vCenter dedicated admin accounts that are not reused for general IT or domain administration, put them behind privileged access management, and require phishing-resistant multi-factor authentication (MFA), a second login factor that cannot be captured by a fake page, on vCenter single sign-on.
- Cut the attack surface. Disable or tightly restrict ESXi SSH and vCenter root SSH, and enable ESXi lockdown mode, which blocks direct host logins and forces management through vCenter. Put the management interfaces on an isolated network reachable only through a hardened jump host.
- Get the logs off the boxes. ESXi writes its logs to a ramdisk that is cleared on every reboot, so local logs vanish when a host restarts. Forward ESXi and vCenter logs to a remote syslog server or a security information and event management (SIEM) system, the platform that collects and alerts on logs. Alert on root SSH logins to vCenter, on Host Profiles changes, on new admin logins, and on AV being disabled on backup servers. Watch for MeshAgent and Ligolo-ng as indicators.
- Put at least one backup out of reach. Keep an immutable copy, one that cannot be altered or deleted before it expires, and keep it apart from the vCenter and domain trust so it cannot be reached with vCenter or domain-admin credentials. Monitor the backup servers for tampering, and test restores on a schedule so you know recovery works before you need it.
The wider lesson
Hypervisor-layer attacks driven by stolen admin credentials walk straight past a patching programme. There was no vulnerability to fix here and no update that would have helped. What decides the outcome is whether one set of stolen credentials can reach both the management plane and the backups. When vCenter is a separate trust boundary and a recovery copy sits outside it, the same intrusion ends with a restore instead of a total loss.
If you run VMware, the work splits into two tracks. Our safeguarding and hardening team locks down vCenter and ESXi management and isolates the backups, and our security operations team watches for the early signs, the root SSH logins, the Host Profiles changes and the AV going quiet on backup servers. Open the chat and Yaali, our AI agent, will pass your question to the engineer who would do the work.
Sources: Kaspersky Securelist, GBHackers, CyberPress, SentinelOne on BlackNevas.
Read next
- Maersk breach shows the risk in acquired systems
- HYFLOCK ransomware's encryption bug and its limits
- Vanderbilt hit by BYOD: check endpoints, Box and OneDrive
Back to the blog, or tell us about your system in the chat. Yaali, our AI agent, answers first and brings in an engineer.