Join our Newsletter — 33% off our NHI Course
Home› Guides› Mustang Panda Threat Actor Profile Guide: PlugX, TONESHELL…
Guide Threat Intelligence & Threat Actors

Mustang Panda Threat Actor Profile Guide: PlugX, TONESHELL and USB campaigns, and how to hunt them

← All guides
By NHI Mgmt Group, edited by Lalit Choda Updated 27 October 2026 11 min read
On this page

Mustang Panda is a China-based espionage group that keeps reaching government, diplomatic and civil society networks with the same simple ingredients: a convincing lure, a legitimately signed program that side-loads a malicious DLL, and malware that can spread by USB drive. It has operated since at least 2012 and has one of the largest technique lists of any China nexus group in MITRE ATT&CK v20. This profile is for SOC analysts, threat hunters and security teams in government, NGOs, think tanks and organisations working on Asia policy. It covers the group's many names, its targeting, its malware families (PlugX, TONESHELL, PUBLOAD and their loaders), how it steals and uses domain credentials, notable operations to 2025, and a hunting and defence summary. Group context sits in the Threat Actor Groups Guide.

Key takeaways

  • Side-loading is the signature. Most chains end with a legitimate, often signed, executable loading a malicious DLL next to it (T1574.001), which makes application control and DLL load monitoring the most useful defences.
  • USB still matters. PlugX and the HIUPAN worm replicate through removable media (T1091); in January 2025 the US Justice Department announced the deletion of a Mustang Panda PlugX version from about 4,258 US computers.
  • Once inside, it goes for the domain. Unit 42 documented Mimikatz against LSASS, DCSync and NTDS.dit theft through volume shadow copies in a Southeast Asian government network.
  • Its own traffic pretends to be TLS. TONESHELL and PUBLOAD use fake TLS headers without a real handshake, a pattern that network sensors can hunt.
  • Edge devices and captive portals are newer entry points. GTIG reported UNC6384, likely associated with Mustang Panda, hijacking captive portal checks to push a signed fake plugin update to diplomats in 2025.

Who they are (names and aliases)

ATT&CK tracks the group as Mustang Panda (G0129), a China-based cyber espionage actor active since at least 2012. The US Justice Department stated in January 2025 that court documents show the PRC government paid the group to develop the PlugX version it removed. Vendor names overlap but are not always one-to-one, so treat each as a related cluster.

NameTracked byNotes
Mustang PandaCrowdStrike, ATT&CK, US Justice DepartmentMost widely used name
Twill Typhoon (earlier TANTALUM)MicrosoftNamed alongside Mustang Panda in the DOJ announcement
Stately TaurusPalo Alto Networks Unit 42Attributed the 2021 to 2023 Southeast Asian government intrusions
Earth PretaTrend MicroReports on TONESHELL and PUBLOAD
Hive0154IBM X-ForceDescribed as overlapping with Mustang Panda
TEMP.Hex, UNC6384Mandiant / Google Threat Intelligence GroupGTIG assesses UNC6384 is likely associated with TEMP.Hex
BRONZE PRESIDENT, TA416, RedDelta, Camaro Dragon, FIREANT, Luminous Moth, Red Lich, ClumsyToadSecureworks, Proofpoint, Recorded Future, Check Point, Broadcom and othersAssociated names listed in ATT&CK

Other China nexus groups are covered in the China State Cyber Operations Guide. PlugX and ShadowPad are shared across several groups; see the PlugX and ShadowPad Malware Guide.

Targeting and motivation

  • Victims: ATT&CK lists government, diplomatic and non-governmental organisations, including think tanks, religious institutions and research entities, across the United States, Europe and Asia, with notable activity in Russia, Mongolia, Myanmar, Pakistan and Vietnam.
  • 2025 targeting: IBM X-Force reported Hive0154 lures aimed at government, military and diplomatic staff in the Philippines, the United States and Pakistan, and a USB infection of a major Taiwanese manufacturer in March 2025. GTIG reported diplomats in Southeast Asia targeted by UNC6384 from March 2025.
  • Lures: current political events and regional issues (South China Sea, Balochistan, spoofed National Security Council notes), invoices and legal documents, and European Commission themes in earlier campaigns.
  • Motivation: intelligence collection in support of PRC strategic interests. The group archives and exfiltrates documents rather than encrypting or destroying systems.

Tactics and techniques (MITRE ATT&CK)

All techniques below are mapped to G0129 in ATT&CK v20; the right-hand column is from the government and vendor reports cited.

TacticTechniqueHow it shows up
Initial accessT1566.001 Spearphishing Attachment, T1566.002 Spearphishing LinkArchives and links to Google Drive or Dropbox holding lure documents and executables
Initial access, lateral movementT1091 Replication Through Removable MediaPlugX variants and the HIUPAN worm copy themselves to USB drives in hidden folders
Credential access, collectionT1557 Adversary-in-the-MiddleCaptive portal hijack redirecting to a fake plugin update (GTIG, UNC6384)
ExecutionT1204.002 Malicious FileUser runs a disguised executable or LNK file
Stealth, executionT1574.001 DLLLegitimate signed executable side-loads CLAIMLOADER, CANONSTAGER or PlugX loader DLLs
Defence impairmentT1553.002 Code SigningDownloaders signed with valid certificates; GTIG counted 25 samples from one signer used by several PRC clusters
PersistenceT1547.001 Registry Run Keys, T1053.005 Scheduled TaskRun keys and scheduled tasks for loaders and TONESHELL
Credential accessT1003.001 LSASS Memory, T1003.006 DCSync, T1003.003 NTDSMimikatz and Hdump; NTDS.dit and SYSTEM hive taken from a volume shadow copy
DiscoveryT1087.002 Domain AccountAdFind and NBTscan; wevtutil queries for logon events
Lateral movement, executionT1072 Software Deployment ToolsAbuse of a security product's remote administration agent to run commands on other hosts
Command and controlT1001.003 Protocol or Service Impersonation; T1219.001 IDE TunnelingFake TLS record headers; Visual Studio Code tunnels through a GitHub account
ExfiltrationT1560.001 Archive via Utility; T1567.002 Exfiltration to Cloud Storage; T1052.001 Exfiltration over USBRAR archives uploaded to Dropbox and file sharing services; USB exfiltration by PlugX variants

Malware families folded into this profile

Family (ATT&CK ID)RoleDefender note
PlugX (S0013), including SOGU.SECLong-running modular backdoorShared by many China nexus groups; USB-spreading variant was the target of the 2025 DOJ operation
TONESHELL (S1239)Backdoor with reverse shell, used since at least 2022Fake TLS headers, some variants with TLS 1.3 style bytes
PUBLOAD (S1228)First-stage downloaderSends a victim ID and pulls the second stage
CLAIMLOADER (S1236)Shellcode loader via side-loadingRun key and scheduled task persistence, decoy PDF
HIUPAN (S1230)USB wormHides files on the drive and leaves a visible UsbConfig.exe
STATICPLUGIN (S1238), CANONSTAGER (S1237)Signed downloader and side-loaded launcherUsed in the 2025 captive portal campaign
ShadowPad (S0596)Modular backdoorSeen alongside TONESHELL in the Unit 42 case

How they abuse identities and credentials

  • Domain-wide credential theft: Unit 42 describes repeated Mimikatz runs against LSASS, DCSync from a domain controller, and a volume shadow copy of a domain controller used to take NTDS.dit and the SYSTEM hive. After that, every domain account, including service accounts, must be treated as known. See the Golden Ticket, Silver Ticket and DCSync Guide and the Credential Dumping Guide.
  • Abuse of trusted management agents: the group used a security product's remote administration agent to execute commands and install backdoors on other hosts, so the agent's service identity became a lateral movement channel. See the RMM Tool Security Guide.
  • Logon intelligence: querying Event IDs 4624 and 4672 with wevtutil let the operators see where privileged users logged on, to choose where to dump credentials next.
  • Abuse of code signing trust: signed loaders and downloaders exploit allowlists that trust any valid signature. See the Code Signing Security Guide.
  • Developer tooling as C2: VS Code tunnels authenticate through the operator's GitHub account, so traffic looks like an ordinary developer session to Microsoft and GitHub endpoints.
  • Compromised email accounts: ATT&CK records the group using compromised legitimate email accounts to send spear-phishing, which defeats sender reputation checks.

Notable incidents

DateOperationSource
Q2 2021 to Q3 2023Long-running intrusion into a Southeast Asian government (CL-STA-0044) with TONESHELL, ShadowPad, Cobalt Strike and domain credential theftUnit 42, September 2023
August 2024 to January 2025French-led operation with Sekoia.io; FBI used nine warrants to delete PlugX from about 4,258 US computersUS Justice Department, 14 January 2025
Late 2024 to May 2025Hive0154 campaigns with TONESHELL, PUBLOAD, CLAIMLOADER and the HIUPAN USB worm against the Philippines, US, Pakistan and TaiwanIBM X-Force, 15 May 2025
March 2025 onwardsUNC6384 captive portal hijack against diplomats in Southeast Asia, delivering SOGU.SEC through a signed downloaderGoogle Threat Intelligence Group, 25 August 2025

Detection and defence

Hunting summary

HuntData sourceWhat to look for
Side-loading pairsEDR image load and file creation eventsA signed executable in an unusual folder (for example under C:\ProgramData\ or C:\Users\Public\) loading an unsigned DLL from the same folder
USB worm artefactsEDR, removable media logsHidden directories on removable drives, a lone executable such as UsbConfig.exe, Explorer settings changed to hide extensions and hidden files
Fake TLS beaconsNetwork sensors (Zeek, NDR)TLS Application Data records (17 03 03, or 17 03 04) with no preceding handshake, as IBM X-Force recommends
Credential dumpingSecurity log, EDRLSASS access by unusual processes, vssadmin shadow copies on domain controllers, directory replication requests from non-DC hosts
Log reconnaissanceProcess creation (Event ID 4688 or Sysmon 1)wevtutil querying Security log for 4624 or 4672
Developer tunnelsProxy and EDRcode.exe tunnel on hosts that are not developer workstations
Captive portal redirectionProxy, DNS, browser telemetryConnectivity check traffic redirected to unknown domains followed by an executable download
Archive and uploadEDR, proxyRAR with passwords, curl or FTP uploads to cloud storage from servers

Controls that break the chain

  1. Application control first: deploy WDAC or AppLocker in audit mode on high-risk user groups, review side-loading candidates, then enforce. Do not trust a signature alone; restrict where signed binaries may run from. See the Application Allowlisting Guide.
  2. Removable media policy: block or read-only USB storage for most users, allow approved encrypted devices, and scan all removable media on connection. See the Removable Media and USB Control Guide.
  3. Protect LSASS and domain controllers: enable LSA protection and Credential Guard, alert on shadow copies and replication from non-DC hosts, and keep privileged accounts in a tiered model.
  4. Control management agents: restrict who can push tasks through EDR, antivirus and RMM consoles, require MFA on those consoles, and log every remote command.
  5. Egress control: restrict servers from reaching consumer cloud storage and file sharing services, and block VS Code tunnel endpoints where developers do not need them.
  6. Travelling staff: require always-on VPN and Enhanced Safe Browsing for diplomats and executives on public networks. See the Public Wi-Fi Safety Guide.

Roll back application control by switching the policy back to audit mode, not by removing it; keep audit data so the next enforcement attempt is better informed.

Response priorities

  • Scope removable media: collect and scan USB drives used on affected hosts, or reinfection follows cleanup.
  • Assume domain compromise if NTDS.dit theft or DCSync is found: plan a full reset including service accounts and KRBTGT.
  • Check management consoles for tasks or deployments you did not create.

Metrics

  • Application control coverage: percentage of endpoints in high-risk groups with an enforced policy.
  • USB exceptions: number of users allowed writeable removable storage.
  • Side-loading alert triage time: median time to close an unsigned DLL load alert.
  • LSA protection coverage: percentage of Windows endpoints and servers with LSA protection enabled.

Framework mapping

FrameworkReferenceRelevance
NIST SP 800-53 Rev. 5CM-7 Least Functionality, CM-11 User-installed Software, SI-3 Malicious Code ProtectionApplication control and side-loading prevention
NIST SP 800-53 Rev. 5MP-7 Media UseUSB worm and USB exfiltration
NIST CSF 2.0PR.PS-05, DE.CM-09Prevent unauthorised software; monitor endpoints

Practitioner checklist

  • Load the G0129 technique set into your ATT&CK coverage view and mark gaps for T1574.001, T1091 and T1003.
  • Run a side-loading hunt across C:\ProgramData\ and C:\Users\Public\ for signed EXE plus DLL pairs.
  • Deploy a network rule for TLS application data without a handshake and tune it for a week.
  • Set removable storage to blocked or read-only for users who do not need it.
  • Enforce application control on diplomatic, policy and executive endpoints first.
  • Enable LSA protection and alert on DC shadow copies and non-DC replication.
  • Review who can run remote tasks through security and management agents.
  • Brief travelling staff on fake plugin and update prompts on hotel and conference Wi-Fi.

Standards and references

Related NHI Mgmt Group resources: Threat Actor Groups Guide · PlugX and ShadowPad Malware Guide · USB Drop Attacks Guide · China State Cyber Operations Guide

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