Join our Newsletter — 33% off our NHI Course
Home› Guides› BRICKSTORM Backdoor Guide: Detecting China nexus implants on…
Guide Attack Techniques & Malware

BRICKSTORM Backdoor Guide: Detecting China nexus implants on vCenter and appliances

PDF for newsletter subscribers · free to read online

← All guides
By NHI Mgmt Group, edited by Lalit Choda Updated 27 October 2026 12 min read
On this page

BRICKSTORM is a backdoor built to live where most security tools cannot see: on VMware vCenter servers, ESXi hosts and network appliances that run no EDR agent. PRC state-sponsored and China nexus actors have used it for espionage campaigns with dwell times measured in years, using the virtualisation layer to clone domain controllers, harvest credentials and reach cloud mailboxes. This guide is for virtualisation, identity and SOC teams. It covers the names and clusters behind the malware, its ATT&CK techniques, how the operators abuse service accounts and vCenter credentials, the main public incidents to February 2026, and how to detect and evict it. Hardening of the platform itself is covered in the Hypervisor and Virtualisation Security Guide and is not repeated here.

Key takeaways

  • Appliances without EDR are the hiding place. Google Threat Intelligence Group (GTIG) reports BRICKSTORM is deployed on appliances that do not support EDR, with an average dwell time of 393 days in its 2025 investigations.
  • The hypervisor is a credential store. Operators cloned domain controller, identity provider and secret vault VMs and mounted the copies offline to extract files such as ntds.dit, so no endpoint agent ever ran.
  • Service accounts carried the intrusion. In the incident CISA describes in AR25-338A, the actors moved with two service accounts and then a managed service provider account to reach vCenter.
  • Cloud access follows. Reporting describes Entra ID enterprise applications with mail.read or full_access_as_app scopes, replayed session tokens and new MFA device registrations.
  • Hunt with vCenter logs, not endpoint telemetry. The CISA Sigma rule needs vCenter logs in a SIEM; VM clone events, new local accounts and SSH enablement are the high-value signals.

Who they are (names and aliases)

BRICKSTORM is malware, not a group. MITRE ATT&CK v20 tracks it as software S9015, a cross-platform backdoor with variants written in Go and Rust, and a variant built from a .NET application using ahead-of-time compilation. ATT&CK lists it as used by several PRC state-nexus clusters. Vendors do not agree that these clusters are one actor, so treat the names as overlapping rather than identical.

NameTracked byRelationship to BRICKSTORM
UNC5221Mandiant / GTIGMain user in GTIG reporting from April 2024. GTIG notes UNC5221 is often used synonymously with Silk Typhoon but does not currently consider them the same
UNC6201Mandiant / GTIGSuspected PRC-nexus cluster with notable overlaps with UNC5221; deployed BRICKSTORM and its successor GRIMBOLT via a Dell RecoverPoint zero-day
WARP PANDACrowdStrikeChina-nexus adversary active since at least 2022, using BRICKSTORM plus ESXi and guest VM implants
PunyToad, SYLVANITEListed in ATT&CK (Cloudflare, Dragos)Other names ATT&CK associates with BRICKSTORM use
PRC state-sponsored cyber actorsCISA, NSA, Canadian Centre for Cyber SecurityAttribution in the joint report AR25-338A

Related tooling in the same reporting: SLAYSTYLE, a JSP web shell (tracked by MITRE as BEEFLUSH); BRICKSTEAL, a malicious Java servlet filter in vCenter; GRIMBOLT, a C# backdoor compiled with Native AOT that replaced BRICKSTORM on some appliances from September 2025; and CrowdStrike's Junction (ESXi) and GuestConduit (guest VM) implants. For the group context see the Silk Typhoon Threat Actor Profile Guide and the China State Cyber Operations Guide.

Targeting and motivation

  • Sectors: AR25-338A names Government Services and Facilities and Information Technology. GTIG adds legal services, SaaS providers, business process outsourcers and technology companies; CrowdStrike lists US legal, technology and manufacturing organisations.
  • Platforms: VMware vCenter and ESXi, Linux-based appliances and, in later reporting, Windows variants. Dell RecoverPoint for Virtual Machines was exploited as a zero-day (CVE-2026-22769, which GTIG scores CVSS 10.0 and says was exploited since at least mid-2024).
  • Objectives: GTIG assesses the operators sought information on US national security and international trade from law firms, access to downstream customers from SaaS providers (high confidence), and intellectual property from technology firms that could support zero-day development.
  • Long dwell by design: GTIG notes that log retention often expired before investigators could find the initial intrusion, and CISA's engagement saw access from April 2024 to at least 3 September 2025.

Tactics and techniques (MITRE ATT&CK)

Techniques below are mapped to S9015 in ATT&CK v20 or described in the government and vendor reports cited. Defensive detail only; no indicators beyond those needed to understand the behaviour.

TacticTechniqueHow it shows up
Initial accessT1190 Exploit Public-Facing ApplicationEdge appliances: Ivanti Connect Secure (CVE-2023-46805, CVE-2024-21887), F5 BIG-IP (CVE-2023-46747), vCenter (CVE-2024-38812, CVE-2023-34048, CVE-2021-22005), Dell RecoverPoint (CVE-2026-22769)
PersistenceT1505.003 Web ShellWeb shell on a DMZ web server (AR25-338A); SLAYSTYLE on vCenter and RecoverPoint
PersistenceT1543 Create or Modify System Process; T1574.007 Path Interception by PATH Environment VariableModified init, rc.local or systemd files; copy in /etc/sysconfig/; PATH changed so the implant runs first
StealthT1036.005 Match Legitimate Resource Name or LocationMasquerades as vCenter processes such as vami-http and updatemgr
StealthT1564.006 Run Virtual Instance; T1070.006 TimestompRogue or unregistered VMs hidden from the vCenter console; timestomping reported by CrowdStrike
Defence impairmentT1690 Prevent Command History LoggingCommand output sent to /dev/null
Credential accessT1556 Modify Authentication ProcessBRICKSTEAL servlet filter on /web/saml2/sso/* captures vCenter sign-in credentials
Credential accessT1003.003 NTDS; T1552.004 Private KeysActive Directory database copied from domain controllers and cloned DC VMs; cryptographic keys exported from an ADFS server
Lateral movementT1021.001 RDP; T1021.004 SSHRDP with service account credentials; SSH enabled through the vCenter VAMI or ESXi UI
Command and controlT1071.001 Web Protocols; T1071.004 DNS; T1572 Protocol Tunneling; T1090.001 Internal ProxyWebSockets with nested TLS, DNS over HTTPS to resolve C2, SOCKS tunnelling into internal networks, VSOCK between VMs
CollectionT1114.002 Remote Email CollectionMailbox access through Entra ID applications granted mail scopes

How they abuse identities and credentials

BRICKSTORM operations are identity operations that happen to use a backdoor. Each stage relies on a credential that was over-privileged, long-lived or unmonitored.

  • Service accounts with interactive rights: AR25-338A describes RDP from a DMZ web server to a DMZ domain controller using a service account, a second service account the next day to an internal domain controller, and then an MSP account to reach vCenter. See the Service Account Security Guide and the Compromised MSP Playbook.
  • vCenter as the master key: once on vCenter, operators could clone VMs of domain controllers, SSO identity providers and secret vaults and read their disks offline. GTIG also reports forensic artefacts suggesting all credentials on a Delinea (Thycotic) Secret Server were decrypted.
  • Built-in platform accounts: CrowdStrike observed use of vpxuser, the account vCenter uses to manage ESXi, and GTIG observed temporary local accounts added to BashShellAdministrators and deleted afterwards.
  • Federation keys: exporting ADFS cryptographic keys gives the ability to sign tokens for federated applications, the same class of risk seen in the SolarWinds Golden SAML case.
  • Cloud application permissions: enterprise applications with mail.read or full_access_as_app read any mailbox without touching a user password. See the Entra ID Applications and Consent Guide.
  • Session replay and MFA persistence: CrowdStrike describes session tokens, likely taken from browser files, replayed through BRICKSTORM tunnels, and a new authenticator device registered for persistence.
  • Hardcoded appliance credentials: the RecoverPoint zero-day was a hardcoded Tomcat admin credential that allowed a malicious WAR upload running as root.

Notable incidents

Date publishedSourceWhat it showed
4 April 2024Mandiant, Ivanti post-exploitationUNC5221 moved from Ivanti Connect Secure to vCenter, created VMs and installed BRICKSTORM disguised as vami-httpd. See the Ivanti Connect Secure exploitation analysis
24 September 2025GTIG, BRICKSTORM espionage campaignLegal, SaaS, BPO and technology victims; 393-day average dwell; BRICKSTEAL, VM cloning, mailbox access; open source BRICKSTORM scanner released
4 December 2025, updated 19 December 2025CISA, NSA, Canadian Centre for Cyber Security, AR25-338A11 samples analysed; incident response case from April 2024 to at least September 2025; YARA and Sigma rules
December 2025CrowdStrike, WARP PANDAESXi implant Junction listening on port 8090, GuestConduit on VSOCK port 5555, Microsoft 365 data access in 2025
17 February 2026GTIG, UNC6201 and Dell RecoverPointZero-day CVE-2026-22769; GRIMBOLT replacing BRICKSTORM; temporary "ghost" NICs on VMs and iptables-based single packet authorisation on vCenter

Government reports on BRICKSTORM have been updated more than once; check the current version of AR25-338A before relying on sample counts or rule names.

Detection and defence

What to hunt for

  • VM clone and creation events: review vCenter VPXD logs for unexpected clones of domain controllers, IdPs and vaults; GTIG noted activity concentrated between 01:00 and 10:00 UTC. Look for unregistered or briefly powered VMs and unexpected virtual NICs.
  • Local account and SSH changes: alert on new local vCenter or ESXi accounts, membership changes to BashShellAdministrators, SSH being enabled and SSH logins as root or vpxuser.
  • Appliance egress: hunt for internet traffic from appliance management IPs, including DNS over HTTPS and WebSocket connections to cloud worker or hosting services.
  • Appliance to Windows logons: Type 3 network logons and RDP from appliance or DMZ addresses to Windows servers, using event logs and Windows User Access Logging.
  • File scanning: scan appliance file systems and backups with the CISA YARA rules or Mandiant's BRICKSTORM scanner, which runs without YARA on appliances.
  • Cloud follow-on: audit enterprise applications with mail read permissions, review MailItemsAccessed records, and look for new MFA registrations from unfamiliar locations.
  • Nonstandard listeners: monitor ESXi for unexpected listeners, notably port 8090 outside its legitimate service.

Hardening that removes the advantages

Advantage the operator relies onControlSource
Unpatched vSphere and edge devicesUpgrade vSphere to the latest version; inventory edge devices and monitor their connectionsAR25-338A
Flat DMZ to internal pathsSegment the DMZ; disable RDP and SMB from DMZ to internal networksAR25-338A
Over-privileged service accountsLeast privilege and extra monitoring for service accounts; rotate credentials and API keysAR25-338A, CrowdStrike
Outbound C2 from appliancesBlock unauthorised and external DoH; restrict outbound traffic from vCenter and ESXiAR25-338A, CrowdStrike
Unsigned binaries on ESXiEnable execInstalledOnly; lockdown mode; disable vpxuser shell access on ESXi 8.0 or laterGTIG, CrowdStrike
Password-only vCenter sign-inAccess vCenter only through an identity provider that requires MFACrowdStrike, GTIG
Vaults and DCs as ordinary VMsTreat credential vault servers and identity VMs as Tier 0, on separately administered infrastructureGTIG
Logs kept only on the deviceForward vCenter and ESXi syslog to an external platform with retention beyond a yearCrowdStrike, GTIG dwell time

See the VMware vCenter Server Hardening Guide, the VMware ESXi Hardening Guide and the Edge Device Security Guide.

Response and eviction

  1. Preserve vCenter, ESXi and appliance logs and disk images before change; the operators have redeployed during incident response.
  2. Scan every appliance and backup, not only the first host found.
  3. Assume compromise of every credential on a cloned or reachable DC, IdP or vault: plan a full Active Directory credential reset, including the KRBTGT account twice, and rotation of ADFS signing keys and vault contents.
  4. Rotate service account, MSP and vpxuser related credentials, and remove temporary local accounts.
  5. Revoke cloud sessions, remove unexplained MFA devices and review application consents.
  6. Rebuild compromised appliances from known-good images where integrity cannot be proved, and only then reconnect.

Metrics

  • Appliance log coverage: percentage of vCenter, ESXi and edge appliances forwarding logs off-box.
  • Tier 0 isolation: number of DC, IdP and vault VMs hosted on general-purpose clusters, target zero.
  • Interactive service accounts: count of service accounts allowed RDP or SSH logons.
  • Appliance egress exceptions: number of appliances allowed unrestricted internet access.
  • Patch latency for KEV-listed appliance CVEs: days from KEV addition to remediation.

Framework mapping

FrameworkReferenceRelevance
NIST SP 800-53 Rev. 5SC-7 Boundary Protection, CM-7 Least Functionality, SI-7 Software, Firmware, and Information IntegritySegmentation, appliance egress, unsigned binary prevention
NIST SP 800-53 Rev. 5AC-6 Least Privilege, IA-5 Authenticator ManagementService and MSP accounts, credential rotation
NIST SP 800-53 Rev. 5AU-6, AU-9Review and protection of hypervisor logs
NIST CSF 2.0ID.AM-01, PR.PS-04, DE.AE-02Appliance inventory, log availability, analysis of events

Practitioner checklist

  • Inventory every vCenter, ESXi host and edge appliance with its management IP, owner and software version.
  • Scan appliances and backups with the CISA YARA rules or the Mandiant scanner, and record the result.
  • Forward vCenter and ESXi logs to the SIEM and deploy the CISA Sigma rule and clone, SSH and local account alerts.
  • Move domain controller, IdP and vault VMs onto separately administered Tier 0 infrastructure.
  • Remove RDP and SMB paths from the DMZ to internal networks and block external DoH.
  • Remove interactive logon rights from service accounts and review MSP access to vSphere.
  • Enforce MFA through an identity provider for vCenter and enable execInstalledOnly on ESXi.
  • Audit Entra ID applications with mail read or full access permissions.
  • Prepare a full credential reset plan that assumes hypervisor-level compromise.

Standards and references

Related NHI Mgmt Group resources: Hypervisor and Virtualisation Security Guide · VMware vCenter Server Hardening Guide · Edge Device Exploitation by Threat Actors Guide · Threat Actor Groups Guide · Service Account Security Guide

PDF for subscribers

Take this guide with you

Every section, table and checklist in one PDF, with links to the sources and related guides. Subscribe to our weekly NHI & AI Identity Journal and we’ll email you a download link.

Free weekly newsletter

Subscribe to the NHI & AI Identity Journal

The latest on NHI and Agentic AI security – articles, research, breaches, news and events every week.

Bonus 33% off our NHI Course when you subscribe.

NHIMG Editorial Note
Researched and written by the NHI Mgmt Group editorial team using AI tools, under the editorial direction of Lalit Choda. Facts are drawn from the sources listed on this page and checked by our automated quality process. Last updated 27 October 2026. Spotted an error? Use our contact page.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM and Agentic AI security. nhimg.org

PDF Downloads for Newsletter Subscribers

Get the guide

Enter your details and we’ll email you a download link.

The full guide is free to read online. The PDF is a thank you for subscribing.

Fill in all fields marked * and tick the PDF box to continue.