BlackNevas ESXi Ransomware Attack Campaign Analysis

An investigative breakdown of how the BlackNevas ransomware group compromised VMware ESXi hosts and vCenter appliances using legitimate management interfaces over a 10-day dwell period.

TL;DR

  • The BlackNevas ransomware group compromised an enterprise virtualization estate over a 10-day period, destroying four ESXi hosts, all hosted virtual machines, two vCenter appliances, and linked backups without using hypervisor zero-day exploits.
  • Attackers established dual-layer management-plane persistence on the vCenter Server Appliance (VCSA) using MeshAgent for initial remote management and a masqueraded Ligolo-ng systemd service (vmware-vdma.service) for network tunneling.
  • Native vCenter Host Profiles were weaponised to remotely enable Tech Support Mode SSH (TSM-SSH) across target ESXi hosts, completing the service state change to RUNNING within 33 seconds.
  • Direct SSH connections to hypervisor hosts originated from the trusted VCSA IP address, leveraging native administrative channels to bypass boundary network controls.
  • Defenders must treat vCenter as a Tier 0 asset, enforce strict credential separation between Active Directory and hypervisors, and actively monitor VCSA systemd units and vSphere management log streams (vpxa.log and vobd.log).

An incident response investigation detailed by Kaspersky Securelist revealed how threat actors executed a destructive campaign using blacknevas esxi ransomware. Over a 10-day dwell period, the adversary systematically destroyed an organisation’s virtualization estate. The scope of destruction encompassed four ESXi hosts, every running virtual machine (VM), both vCenter Server Appliances (VCSA), and the primary backup infrastructure. Rather than relying on hypervisor-layer memory corruption or guest-to-host escapes, the threat actor achieved total estate compromise by executing operations exclusively through valid VMware management interfaces and native vCenter workflows.

The campaign highlights a persistent operational challenge in enterprise infrastructure security: the blind spot created when management planes are compromised. Ransomware operators have increasingly focused their efforts on hypervisor layers. Kaspersky’s Global Emergency Response Team (GERT) reported that ransomware accounted for 39.4% of all incidents handled in 2025. This trend has been sustained by the broad availability of ESXi lockers derived from the leaked Babuk source code, which lowers the technical bar for threat groups seeking host-wide encryption capabilities.

By establishing persistent control over vCenter Server Appliances and leveraging native administrative mechanisms—specifically vCenter Host Profiles—the BlackNevas group bypassed traditional endpoint telemetry entirely. Understanding how native management tools are weaponised is essential for security engineers and system administrators tasked with defending virtualized infrastructure.

Anatomy of a 10-Day Virtualization Estate Compromise

The campaign analyzed by Kaspersky GERT involved an adversary remaining inside the victim network for ten days before launching the final destructive phase. During this dwell time, the adversary mapped the virtualization topology, established redundant persistence mechanisms, and prepared access routes to every hypervisor under the management of the targeted vCenter appliances.

The impact on the victim infrastructure was total across the targeted virtualization assets:

  • Four ESXi Hosts: All targeted bare-metal hypervisors were accessed via administrative channels, leading to host-wide file encryption or destruction.
  • All Hosted Virtual Machines: Workloads including domain controllers, application servers, and databases were rendered unrecoverable at the guest level due to external modification of underlying Virtual Machine Disk (VMDK) files.
  • Dual vCenter Server Appliances: Both primary and disaster recovery vCenter appliances were compromised and used as launching points for administrative tasks.
  • Connected Backup Infrastructure: Backup servers and repositories were targeted directly to prevent fast restoration, neutralizing off-site and local recovery options.

A critical finding from the post-incident investigation was the total absence of hypervisor zero-day vulnerabilities or complex guest-to-host escape primitives. The BlackNevas group, an operational entity first observed in November 2024, did not require memory corruption exploits against VMkernel. Instead, they relied entirely on abused administrative privileges, native utilities, and misconfigured infrastructure trust boundaries.

This operational model relies heavily on the commodotisation of ESXi ransomware lockers. Following the September 2021 leak of the Babuk ransomware source code, including its dedicated ESXi build target, at least ten distinct threat groups adopted Babuk-derived code bases. These lockers utilize standard Linux system calls and POSIX file system APIs to identify and lock .vmdk, .vmem, and .vswp files. By combining these widely available lockers with administrative access to vCenter, threat actors eliminate the need to develop proprietary hypervisor exploits.

Undisclosed Incident Details

According to the Kaspersky report, the exact initial entry vector used to secure the initial foothold inside the victim organisation’s broader enterprise network was not disclosed. Additionally, specific victim identity data and the precise internal file-wiping commands executed against backup datastores remain unconfirmed in public telemetry.

The Visibility Asymmetry in Hypervisor Environments

Securing a hypervisor estate requires addressing a fundamental architectural challenge: traditional security monitoring paradigms designed for general-purpose operating systems break down at the VMkernel layer. This creates a visibility asymmetry that adversaries systematically exploit.

ESXi is built on VMkernel, a proprietary microkernel operating system designed specifically for running virtual machines. While VMkernel presents a POSIX-like command shell and supports familiar Unix-style commands, its internal scheduler, memory management architecture, and driver model differ fundamentally from standard Linux distributions. General-purpose Endpoint Detection and Response (EDR) agents cannot run natively on VMkernel.

+-------------------------------------------------------------------+
|                        vCenter Server Appliance                   |
|  [ MeshAgent (Remote Access) ]   [ Ligolo-ng Tunnel (Systemd) ]   |
+----------------------------------+--------------------------------+
                                   |
                                   | Abused Management Protocols
                                   | (vpxd / Host Profiles / SSH)
                                   v
+-------------------------------------------------------------------+
|                           ESXi Host                               |
|  +-------------------------------------------------------------+  |
|  |                     VMkernel Hypervisor                      |  |
|  +-------------------------------+-----------------------------+  |
|                                  | Direct File Modification       |
|                                  v (Outside Guest Boundary)       |
|  +-------------------------------------------------------------+  |
|  |                 Virtual Machine Datastore                   |  |
|  |  +-----------------------+       +-----------------------+  |  |
|  |  | Guest VM (Windows)    |       | Guest VM (Linux)      |  |  |
|  |  | [ Guest EDR Active ]  |       | [ Guest EDR Active ]  |  |  |
|  |  |   - Sees IO Errors    |       |   - Sees IO Errors    |  |  |
|  |  |   - No Process Context|       |   - No Process Context|  |  |
|  |  +-----------------------+       +-----------------------+  |  |
|  +-------------------------------------------------------------+  |
+-------------------------------------------------------------------+

Broadcom explicitly states that direct installation of third-party agents, antivirus software, or security monitoring binaries on ESXi hosts and the vCenter Server Appliance is unsupported. Installing unapproved agents inside the ESXi shell risks destabilising the hypervisor, invalidating support agreements, and inducing storage latency.

This constraint creates a telemetry gap between the guest operating system and the hypervisor host:

  • Guest EDR Limitation: EDR sensors installed inside a guest VM rely on kernel drivers (such as Windows sysmon or Linux eBPF probes) to inspect local processes, registry modifications, and file operations.
  • External Datastore Modification: When an adversary accesses the underlying ESXi host directly via SSH or management APIs and initiates encryption against raw .vmdk files, the operation occurs entirely outside the guest kernel context.
  • Guest Symptoms: The guest EDR records symptoms—such as abrupt read/write timeouts, file system read-only remounts, or system crashes—but cannot capture the binary path, process ID, command line, or network context of the hypervisor-level process causing the destruction.

Because third-party agents cannot be deployed to VMkernel, defenders must rely on infrastructure-level audit logs, out-of-band management plane logging, agentless security platform integrations, and centralized syslog collection from vpxa.log, hostd.log, and vobd.log.

Entry Vectors: From Active Directory Abuse to Credential Bridges

Adversaries target four main entry routes to reach the virtualization management plane and compromise ESXi infrastructure:

  1. Unpatched management-plane vulnerabilities
  2. Active Directory identity integration abuse
  3. Credential reuse from compromised administrative endpoints
  4. Backup platform credential extraction

Vulnerability Exploitation in Management Systems

Although vulnerability exploitation was not identified as the access vector in the BlackNevas engagement, public vulnerabilities in vCenter and ESXi represent high-frequency entry paths for ransomware operators across the industry. The table below outlines key vulnerabilities affecting hypervisor management infrastructure documented in threat telemetry.

CVE ID Affected Layer Security Impact CISA KEV Status Threat Context & Relevance
CVE-2021-21972 vCenter Plugin Unauthenticated RCE Yes Chained by threat groups (e.g., Twelve) to deploy web shells directly on vCenter.
CVE-2021-21974 ESXi OpenSLP Remote Code Execution No Historically leveraged in widespread ESXiArgs ransomware campaigns.
CVE-2021-22005 vCenter Server Arbitrary File Upload / RCE Yes Widely exploited to obtain initial shell access on VCSA instances.
CVE-2021-21985 vCenter vSAN Plugin Remote Code Execution Yes Documented use in automated management-plane exploitation chains.
CVE-2023-34048 vCenter DCERPC Out-of-Bounds Write RCE Yes Critical unauthenticated vulnerability affecting vCenter management services.
CVE-2024-37085 ESXi AD Integration Domain Group Privilege Escalation Yes Exploited by Akira and Black Basta to map domain admin rights directly to ESXi root.
CVE-2024-38812 vCenter DCERPC Remote Code Execution Yes Heap overflow in vCenter DCERPC protocol handling.
CVE-2025-22225 ESXi / Workstation Guest-to-Host Escape Chain Yes Requires prior privileged execution inside a guest VM to compromise host.

Active Directory Abuse and CVE-2024-37085

Active Directory integration presents a critical trust boundary issue. CVE-2024-37085 illustrates how domain identity misconfigurations can instantly translate into hypervisor compromise. On unpatched ESXi hosts joined to an Active Directory domain, any domain user account added to a group named ESX Admins (by default) is automatically granted full administrative privileges on the ESXi host.

Exploitation by groups such as Akira and Black Basta involves creating the ESX Admins group within Active Directory if it does not exist, assigning target accounts to it, and leveraging the automated permission synchronization to log in to ESXi hosts as root equivalents.

Broadcom remediated this issue in ESXi 8.0 U3. For legacy or unpatched systems, defenders must manually execute configuration workarounds to break this inheritance:

# Set Advanced Configuration parameters via ESXi Command Line (test in staging)
# Disable automatic group provision for ESX Admins
esxcli system settings advanced set -o /UserVars/esxAdminsGroupAutoAdd -b false

# Reduce authentication validation interval from default (1440 mins) to 90 mins
esxcli system settings advanced set -o /UserVars/authValidateInterval -i 90

# Re-validate configured AD Security Group settings
esxcli system settings advanced list -d | grep -i esxadmins

Credential Reuse and Backup Bridges

Adversaries often gain access without triggering vulnerability exploits by harvesting existing credentials. When system administrators reuse credentials across domain controllers, jump hosts, and vCenter management portals, an initial compromise in the Windows estate quickly spills over to the hypervisor plane.

Furthermore, backup infrastructure functions as a concentrated credential bridge. Backup solutions require high-privilege service accounts, management SSH keys, and broad network access to interact with ESXi hosts, store VM snapshots, and manage storage LUNs. In the BlackNevas engagement, the adversary obtained administrative execution on the backup infrastructure, extracted stored virtualization credentials, and pivoted directly to the vCenter appliances.

Management-Plane Persistence: MeshAgent and Ligolo-ng

Once initial root access to the targeted vCenter Server Appliance was established via SSH, the BlackNevas operators established dual-layer persistence to ensure continuous access to the management plane, even if interactive SSH access was later restricted.

The attackers executed a deliberate, multi-step persistence sequence:

  1. Initial SSH Access: The adversary authenticated to the VCSA via SSH using compromised root credentials.
  2. Rapid Utility Deployment: Exactly 43 seconds after authenticating, the adversary executed a shell script to fetch and install MeshAgent—the lightweight agent component of the open-source MeshCentral remote management platform.
  3. Session Termination: The initial interactive SSH session was terminated after a total duration of 2 minutes and 26 seconds. The threat actor did not establish subsequent interactive SSH sessions to the VCSA, neutralizing defenders relying solely on SSH session logging.
  4. Secondary Persistence Installation: Approximately four hours later, the adversary re-engaged through the active MeshAgent connection to install Ligolo-ng (version 0.8.2), an advanced TUN/TAP interface protocol tunneling framework.

Masquerading as Native VCSA Services

To avoid manual detection by sysadmins reviewing active system services, the adversary registered Ligolo-ng as a persistent systemd service using legitimate-looking VMware naming conventions and paths.

The malicious systemd unit file was placed in the system service directory, masquerading as a native daemon:

# Artifact Location: /etc/systemd/system/vmware-vdma.service
[Unit]
Description=DNS, time, VM settings sync (v2.03.24766)
After=network.target

[Service]
Type=simple
ExecStart=/opt/vmware/bin/agent
Restart=always
RestartSec=5s

[Install]
WantedBy=multi-user.target

String analysis of the compiled binary located at /opt/vmware/bin/agent revealed its true identity as Ligolo-ng v0.8.2, configured to establish an outbound encrypted C2 tunnel to a remote endpoint: api.cloudvflare[.]com:443.

This persistence tactic maps directly to MITRE ATT&CK techniques:

  • T1543.002 (Create or Modify System Process: Systemd Service): Used to maintain boot persistence across appliance reboots and automatically restart the tunnel process if terminated.
  • T1572 (Protocol Tunneling): Used to encapsulate arbitrary internal management network traffic inside an outbound TLS connection over port 443, bypassing inbound firewall restrictions.

Defenders can audit systemd services on vCenter appliances by querying for unmanaged or non-standard unit files using systemctl:

# List installed systemd unit files on VCSA filtering for non-standard paths
systemctl list-unit-files --type=service | grep vmware

# Inspect unit file paths outside standard distribution directories
ls -la /etc/systemd/system/vmware-*.service

# Verify digital signature or package ownership of binaries under /opt/vmware/bin/
rpm -qf /opt/vmware/bin/agent

Weaponising vCenter Host Profiles for Lateral Movement

The most distinctive technical element of the BlackNevas campaign was how the group leveraged vCenter Host Profiles to orchestrate lateral movement from the VCSA to individual ESXi hypervisors. Host Profiles are designed for legitimate cluster management: they allow administrators to capture a baseline configuration from a reference host and push those configuration policies to multiple managed hosts to ensure consistency.

Host Profiles include configuration modules for the host services layer, specifically managing Tech Support Mode SSH (TSM-SSH). Under hardened baselines, TSM-SSH is kept in a STOPPED state to prevent unauthorized remote shell access.

The adversary altered the Host Profile applied to the cluster, changing the TSM-SSH service configuration policy from policy-disabled to enabled. Re-applying the Host Profile pushed this configuration change to all managed ESXi hosts simultaneously.

+-------------------------------------------------------------------+
|                     vCenter Server Appliance                      |
|                                                                   |
|   1. Read Host Configuration (T+0)                                |
|   2. Push Modified Host Profile (T+10m18s)                        |
|      - Set TSM-SSH policy to ENABLED                              |
+---------------------------------+---------------------------------+
                                  |
                                  | vpxa RPC / Host Profile Push
                                  v
+-------------------------------------------------------------------+
|                      Target ESXi Host                             |
|                                                                   |
|   3. TSM-SSH state changes to RUNNING (T+10m51s)                  |
|      - Logged in vpxa.log and vobd.log                            |
|                                                                   |
|   4. Incoming Direct SSH Connection (T+11m+)                      |
|      - Source IP: vCenter Server Appliance IP                     |
+-------------------------------------------------------------------+

Log correlation across the management plane captured the exact operational sequence:

Relative Time Log Source Recorded Event Description Technical Significance
T+0 vpxa.log (vCenter) Host configuration read via Host Profile workflow Adversary inspects current baseline configuration across managed hosts.
T+10m18s vpxa.log (vCenter) Host profile configuration push initiated State change sent from vCenter to target ESXi host agent (vpxa).
T+10m51s vpxa.log / vobd.log TSM-SSH service reached RUNNING state SSH service successfully enabled and listening on TCP port 22.

The interval between the host profile push (T+10m18s) and the TSM-SSH service reaching the RUNNING state (T+10m51s) was exactly 33 seconds. Eleven minutes after initial configuration inspection, the adversary had reconfigured the cluster to support direct remote SSH access.

Crucially, when the adversary subsequently initiated SSH connections to the ESXi hosts, the incoming traffic originated from the internal IP address of the vCenter Server Appliance. Because network firewalls and ESXi host firewalls are routinely configured to trust all traffic from their managing vCenter appliance, these malicious administrative connections bypassed perimeter and internal network filters.

Hardening the Virtualization Trust Stack and Detection Engineering

Defending against management-plane exploitation requires structural adjustments to virtualization architecture, combined with high-fidelity logging strategies focused on administrative configuration changes.

Classifying vCenter as a Tier 0 Asset

vCenter Server Appliances control access to all underlying hypervisors, virtual storage, and hosted workloads. Consequently, vCenter must be classified as a Tier 0 asset—subject to the same security controls, network isolation, and access restrictions as Active Directory Domain Controllers.

  • Dedicated Management Networks: Place vCenter appliances and ESXi management interfaces on isolated, out-of-band management VLANs. Access must be restricted to dedicated management jump boxes via strict network access control lists (ACLs).
  • Identity Separation: Do not use primary Active Directory domain admin accounts for vCenter administration. Enforce dedicated local vCenter accounts or isolated identity federation systems using strict multi-factor authentication (MFA).
  • Restrict SSH Shell Access: Disable SSH on both vCenter appliances and ESXi hosts by default. Enable interactive shell access only during explicitly scheduled change windows.

Log Correlation and Telemetry Ingestion

Because third-party agents cannot run natively on VMkernel, security teams must configure centralized syslog forwarding from ESXi hosts and vCenter appliances to an external Security Information and Event Management (SIEM) system. Key log files include:

  • /var/log/vpxa.log: Records interaction between vCenter Server and the ESXi host agent.
  • /var/log/hostd.log: Captures host-level management tasks, direct ESXi web client connections, and API calls.
  • /var/log/vobd.log: The VMkernel Observation Bearer Daemon log, recording high-level system state changes such as service startup events.
  • /var/log/shell.log: Tracks ESXi shell login events, executed commands, and SSH session activity.

The following example rule demonstrates how defenders can detect unexpected SSH enablement via Host Profiles or direct management calls by monitoring syslog streams for TSM-SSH state changes:

# Example rule logic for detecting ESXi SSH enablement via syslog ingestion
# Rule Logic: Trigger alert if TSM-SSH moves to RUNNING outside approved change window

rule ESXi_SSH_Service_Enabled {
    meta:
        description = "Detects TSM-SSH service state changing to RUNNING"
        severity = "HIGH"
    events:
        $log in /var/log/vobd.log or syslog stream
        
        // Filter for service state change event in VMkernel Observation logs
        ($log contains "vob.user.ssh.enabled" or 
         $log contains "TSM-SSH service state is RUNNING" or
         ($log contains "vpxa" and $log contains "HostProfile" and $log contains "TSM-SSH"))
         
    condition:
        $log
}

Common Hardening Mistakes

Organizations often make critical architecture errors when securing virtualization estates:

  • Implicit Trust of VCSA Traffic: Assuming that connections originating from the vCenter IP address are inherently safe, leaving hypervisors exposed if vCenter is compromised.
  • Shared Administrative Credentials: Reusing the same root passwords or administrative domain accounts across hypervisors, backup appliances, and guest operating systems.
  • Unmonitored Systemd Services: Auditing user logins on Linux-based virtual appliances while ignoring unauthorized systemd unit file additions in /etc/systemd/system/.
  • Co-located Storage Backups: Storing VM snapshots and backups on storage arrays that share authentication domains or management networks with the hypervisors they back up.

Key Takeaways for Infrastructure and Security Leaders

The BlackNevas campaign demonstrates that threat actors can completely destroy virtualized estates without using sophisticated hypervisor zero-day exploits. Managing this risk requires structural updates to enterprise defensive architectures.

Action Items for Infrastructure Teams

Isolate hypervisor management interfaces, enforce strict credential boundaries, and continuously audit VCSA system units and Host Profile baseline changes.

  • Decouple Hypervisor Identities from Active Directory: Eliminate direct Active Directory administrative dependencies on ESXi hosts wherever possible. Where AD integration is required, apply Broadcom’s workarounds for CVE-2024-37085, disable esxAdminsGroupAutoAdd, and strictly restrict domain group membership.
  • Address the Hypervisor Telemetry Gap: Recognize that guest-level EDR cannot observe hypervisor-level storage manipulation. Centralize syslog export for vpxa.log, vobd.log, and hostd.log to maintain audit visibility across all hypervisors.
  • Enforce Immutable, Air-Gapped Backups: Protect backup infrastructure by isolating management interfaces and configuring write-once-read-many (WORM) storage targets to ensure data recovery if management planes are compromised.
  • Audit Virtual Appliance Integrity: Regularly monitor vCenter Server Appliances for unauthorized software, modified systemd service descriptors (such as /etc/systemd/system/vmware-*.service), and unexpected outbound TLS connections over non-standard management channels.

Frequently asked questions

Why can traditional EDR agents not be installed directly on ESXi hosts?

Broadcom explicitly does not support installing third-party EDR software or agents directly on ESXi or vCenter Server Appliances. ESXi runs on VMkernel, a proprietary microkernel OS with unique memory management and process scheduling models. Standard Linux security tools can destabilise hypervisor performance and void vendor support agreements.

How do ransomware operators compromise ESXi without exploiting hypervisor vulnerabilities?

Threat actors gain access to vCenter appliances or ESXi hosts using valid administrative credentials obtained through credential reuse, phishing, Active Directory integration abuse, or compromised backup servers. Once authenticated as root or administrator, they execute host-wide encryption using native management interfaces and command shells without relying on zero-day vulnerabilities.

What role did vCenter Host Profiles play in the BlackNevas campaign?

The BlackNevas threat group weaponised vCenter Host Profiles to modify cluster-wide compliance settings. They altered the profile to enable Tech Support Mode SSH (TSM-SSH) and pushed the updated configuration across managed ESXi hosts, automatically opening SSH ports host-wide within 33 seconds to enable remote administrative access.

How can defenders detect malicious systemd services on vCenter Server Appliances?

Defenders can audit systemd services by connecting to the VCSA shell and checking for unit files created outside default distribution paths, particularly in /etc/systemd/system/. Commands such as systemctl list-unit-files --type=service help identify suspicious, newly created, or unverified services disguised with VMware-like naming conventions.

On legacy or unpatched ESXi systems where upgrading to version 8.0 U3 is not immediately possible, defenders should set the advanced configuration setting /UserVars/esxAdminsGroupAutoAdd to false using esxcli. Additionally, lower the authentication validation interval (/UserVars/authValidateInterval) to reduce exposure to unauthorized domain administrative group mappings.

Sources and further reading

This article was drafted with AI assistance from the publicly cited sources and passed automated checks for structure and sourcing. Verify critical details against the vendor advisories linked above before acting.