Attackers gained administrative access to VMware vCenter servers through an authentication bypass that has now been patched. Forensic evidence indicates a suspected China-linked group used the weakness in early September after first compromising BeyondTrust Remote Support and moving through internal Linux appliances. The attackers created a vCenter single sign-on administrator account, stole directory data and obtained material that could support forged authentication tokens. Broadcom classifies the fix as an emergency update, and no official workaround is available.

What Happened

Broadcom released an emergency update for a VMware vCenter Server authentication bypass after investigators documented exploitation dating to early September 2026. The weakness affects the VMware Directory Service in vCenter, the centralized platform used to configure, control and monitor ESXi environments. Successful exploitation can create an authenticated session as administrator@vsphere.local without requiring the attacker to authenticate normally.

The observed activity was attributed to ARMORY CASTLE, a suspected China-linked threat actor formerly tracked as UNC5221. In the documented intrusion, the group first compromised a BeyondTrust Remote Support appliance through a separate zero-day vulnerability. It then moved laterally to internal Linux appliances, escalated privileges to root and searched the environment for VMware systems.

The attackers focused reconnaissance on VMware-specific services and deployed multiple web shells, backdoors and utilities. One custom Golang tool, BONESAW, exploited the vCenter weakness to create a single sign-on administrator account and extract the server’s directory database. The tool could also obtain the Security Token Service signing key and support forged administrative SAML assertions.

A second utility, SAWSNAKE, used material from the vCenter database to forge authentication tokens, bypass multifactor authentication and impersonate users. Reporting on the intrusion said the attackers moved deeper into the virtualized environment, reached ESXi hosts and extracted sensitive information from virtual machines.

The campaign used two zero-day vulnerabilities and several malware families tailored to the targeted platforms. Broadcom patched the vCenter issue on November 19. It said affected organizations should install the fixed versions as soon as possible. There is no official workaround, although network restrictions and firewall controls may reduce exposure.

Why It Matters

vCenter is a high-value control plane. Administrative access can give an attacker influence over the virtual machines managed through the server, making compromise more consequential than the loss of one ordinary endpoint. The observed theft of directory data and signing material also means defenders must consider identity trust, not just malware removal.

The intrusion path is important because the attackers did not begin at vCenter. They entered through a remote-support product, moved through internal Linux appliances and then targeted virtualization management. Some of those Linux systems lacked endpoint detection and response tooling. This demonstrates how appliances with limited monitoring can become stepping stones to tier-zero infrastructure.

The ability to forge SAML tokens is especially serious because it can undermine normal authentication controls, including multifactor authentication. Resetting a password alone may not resolve that form of compromise. Organizations need to validate administrative accounts, token-signing trust and the integrity of the management environment.

For executives, this is a concentration-of-risk issue. A centralized virtualization platform can support many important workloads, so patch delays or weak segmentation can create a broad business exposure. For defenders, the case supports treating virtualization administration, remote access and identity-signing systems as a connected security boundary.

What to Watch

Organizations should watch for further technical guidance from Broadcom and updated detection material from the investigators. The evidence already establishes exploitation, but additional indicators may improve hunting across vCenter, remote-support appliances and intermediate Linux systems.

Defenders should determine whether affected vCenter servers were reachable from systems that could have been compromised, even if the servers were not directly exposed to the internet. The documented campaign reached vCenter through internal lateral movement, so perimeter exposure alone is not a sufficient measure of risk.

Any discovery of an unauthorized single sign-on administrator, unexpected directory database access or suspicious SAML activity should be treated as potentially significant. Because stolen signing material can affect trust, responders may need a broader recovery decision than simply deleting visible malware. Actions should follow vendor and incident-response guidance.

The full scope of affected victims was not provided in the supplied evidence. Organizations should therefore avoid assuming that absence of public victim reporting means absence of targeting.

What Defenders Should Do

Emergency patching: Apply Broadcom’s fixed versions immediately across affected vCenter deployments and related products covered by the advisory. There is no official workaround.

Access reduction: Restrict network access to vCenter and its supporting services to explicitly authorized management systems. Firewall controls can reduce exposure but should not replace patching.

Threat hunting: Investigate for unexpected single sign-on administrator accounts, unusual access to the directory database, suspicious command execution and evidence of SAML token forgery. Review the intermediate Linux appliances and remote-support systems that can reach vCenter.

Visibility: Ensure endpoint or equivalent monitoring covers critical Linux appliances where supported. Centralize logs from remote-access, identity and virtualization systems so investigators can reconstruct lateral movement.

Recovery: If compromise is suspected, assess whether directory data or signing keys were accessed. Use vendor and incident-response guidance to restore identity trust rather than relying only on password rotation or malware deletion.

The Larger Lesson

Attackers increasingly target the systems that administer other systems. Remote-support appliances, identity services and virtualization control planes form a chain of trust, and compromise at one point can create a path to the next.

The durable lesson is to manage these platforms as tier-zero infrastructure. They need rapid patching, strict network paths, strong logging, independent recovery plans and monitoring that extends beyond standard employee endpoints. A single management foothold can otherwise become broad control over workloads, identities and sensitive data.

Questions, corrections, or feedback about this story?Contact Daily Cyber News →

← Back to this Daily edition