العربية
🚨 BreakingSonicWall Gen 6 Firewalls Reach End of Security Patches

Armory Castle Exploited vCenter CVE-2026-123890 as a Zero-Day

فريق السايبر العربياقرأ بالعربية

Mandiant and Google say Armory Castle used CVE-2026-123890 on vCenter in early September 2026, more than two months before Broadcom's 19 November patch.

Security World illustrative cover: Armory Castle Exploited vCenter CVE-2026-123890 as a Zero-Day
Illustration by Security World

⚡ Why it matters

vCenter is the control plane for vSphere, so an authentication bypass can yield administrative control of virtual infrastructure. This article updates our earlier coverage of Broadcom's advisory with confirmed pre-patch exploitation.

This is a follow-up to our earlier report on the critical VMware vCenter flaw CVE-2026-123890. After Broadcom published its advisory on 19 November 2026, Mandiant and Google Threat Intelligence Group said the bug had already been used as a zero-day since early September 2026 — more than two months before a fix was available — according to SecurityWeek.

Who exploited it

Mandiant, which reported the issue to Broadcom, attributes the activity to the espionage group Armory Castle, previously tracked as UNC5221. Google describes the group as a suspected PRC-nexus espionage actor. SecurityWeek adds that the group is also known as Warp Panda. Charles Carmakal, Mandiant's CTO, said on LinkedIn that the firm investigated multiple incidents in which the group exploited this vulnerability, SecurityWeek reported.

How the intrusion unfolded

According to Mandiant and SecurityWeek, the operators first gained access through CVE-2026-102764, a recently patched zero-day in BeyondTrust Remote Support, then moved to Linux appliances on the internal network and targeted vCenter in early September.

They used a custom Golang utility that Mandiant tracks as BONESAW. The tool exploits CVE-2026-123890 to bypass authentication and open a session as [email protected], then steal the Lightning Memory-Mapped Database (LMDB), extract SAML signing material, and create a single sign-on (SSO) administrator account.

The group then deployed a second utility, tracked by Mandiant as SAWSNAKE, to forge SAML tokens and generate administrative sessions. They also obtained the vCenter master encryption key and database credentials, moved to an ESXi host, and took additional data. In the intrusion Mandiant analyzed, the actors also extracted files from guest systems on that host.

Mandiant says BONESAW achieves full compromise of vCenter 7.x and 8.x appliances. The firm also reports evidence that Armory Castle knew about CVE-2026-123890 before the earlier directory-service bypass CVE-2026-59309 was disclosed on 29 July 2026, and that the tool was used against a vCenter server already patched for that earlier bug.

Fixed versions

Broadcom shipped fixes in advisory VMSA-2026-0009 for vCenter 9.1.1.0200, 9.0.2.0200 and 8.0 U3m. There is no public patch for version 7.0, which reached end of general support in October 2025; customers with an extended support contract can contact Broadcom Support. SecurityWeek notes that Broadcom's own advisory does not mention in-the-wild exploitation.

Recommendations

  • It is recommended to update to 9.1.1.0200, 9.0.2.0200 or 8.0 U3m, depending on the supported branch, or to request a fix from Broadcom if running 7.0 under extended support.
  • It is recommended to inspect vCenter systems for signs of compromise, including unexpected SSO administrator accounts, SSH being turned on through the VAMI interface without an operational reason, and inbound connections to TCP port 2012 from hosts that should not reach the directory service.
  • It is recommended to restrict network access to vCenter and keep it isolated from the internet and untrusted internal systems.
  • It is recommended to confirm that the fix for CVE-2026-102764 is applied on BeyondTrust Remote Support deployments, where that product is in use.

Sources: SecurityWeek — Mandiant / Google Threat Intelligence Group — Broadcom VMSA-2026-0009

🔗 Sources

This article is original reporting by the Security World editorial team based on the sources listed above; direct quotes are kept short and marked with quotation marks. Corrections and feedback: info [at] securitywd.net.

Related stories