Join our Newsletter — 33% off our NHI Course
Home› Guides› APT5 Threat Actor Profile Guide: VPN and ADC…
Guide Threat Intelligence & Threat Actors

APT5 Threat Actor Profile Guide: VPN and ADC appliance attacks and defences

PDF for newsletter subscribers · free to read online

← All guides
By NHI Mgmt Group, edited by Lalit Choda Updated 27 October 2026 13 min read
On this page

APT5 is a China-based espionage group that goes after the appliances sitting at the edge of a network, the SSL VPNs and application delivery controllers that every remote user and federated login passes through. MITRE ATT&CK tracks it as G1023 and describes activity since at least 2007 against telecommunications, aerospace and defence organisations in the United States, Europe and Asia, with a marked interest in networking devices and the use of zero-day exploits. In December 2022 the US National Security Agency (NSA) published threat hunting guidance for Citrix Application Delivery Controller (ADC) after APT5 activity against those devices, and Mandiant's 2021 Pulse Secure investigations described a related cluster, UNC2630, that bypassed single and multi-factor authentication on VPN appliances. This profile is for SOC analysts, edge device owners and threat intelligence teams. It covers names and attribution as each source states it, targeting, ATT&CK techniques, how the group abuses credentials and authentication, notable events and a detection and hardening plan. The wider picture of state actors exploiting edge devices is in the Edge Device Exploitation by Threat Actors Guide; this page covers what is specific to APT5.

Key takeaways

  • The appliance is the target, not a stepping stone you can ignore. APT5 has exploited Pulse Secure (now Ivanti) Connect Secure VPNs and Citrix ADC, and has modified appliance binaries and upgrade scripts so its access survives patching.
  • Authentication is bypassed at the source. Mandiant found trojanised Pulse Secure libraries that forced a successful LDAP, RADIUS or RSA ACE result, including the second factor, when a hardcoded password was supplied.
  • Harvested credentials carry the intrusion inward. Credential logging on the VPN, LSASS and SAM dumping on Windows and stolen credentials reused against Microsoft 365 turn one appliance into domain and cloud access.
  • Logs on the device cannot be trusted. Tools such as THINBLOOD and CLEANPULSE wiped or suppressed appliance log entries, so detection depends on off-device logging and correlation with the identity provider.
  • Integrity checking is the decisive control. NSA and Mandiant both lead with comparing appliance binaries to known-good hashes, running the vendor integrity tool and restoring compromised devices to a known-good state rather than cleaning them in place.

Who they are (names and aliases)

Vendors and governments name the group differently, and some names describe overlapping clusters rather than an identical set of operators. The table records each name with the source that uses it.

NameUsed byNotes
APT5Mandiant, MITRE ATT&CK (G1023), NSAThe primary name in ATT&CK; China-based espionage actor active since at least 2007.
Mulberry TyphoonMicrosoftMicrosoft's current name; its naming table lists MANGANESE, KEYHOLE PANDA and APT5 among the other names.
MANGANESEMicrosoft (former naming scheme), NSAThe NSA Citrix ADC advisory refers to "APT5 (also known as UNC2630 and MANGANESE)".
Keyhole PandaCrowdStrike naming, listed by ATT&CK and MicrosoftAlias only; no separate reporting used here.
BRONZE FLEETWOODSecureworks, listed by ATT&CKAlias only.
UNC2630MandiantPulse Secure intrusion cluster. Mandiant suspected a link to APT5 but said the link was not definitive; ATT&CK lists it as an associated group.

Attribution caution: Mandiant's May 2021 update said a significant number of the organisations compromised through Pulse Secure had previously been APT5 targets, and that multiple groups were likely responsible. Treat UNC2630 activity as closely related to APT5 rather than proof that every Pulse Secure intrusion was APT5. The broader Chinese state context is in the China State Cyber Operations Guide.

Targeting and motivation

  • Sectors: ATT&CK lists telecommunications, aerospace and defence as the primary targets. Mandiant's 2021 investigations covered US Defense Industrial Base companies (August 2020 to March 2021), plus government and financial organisations.
  • Geography: the United States, Europe and Asia, according to ATT&CK.
  • Motivation: espionage. Mandiant assessed the Pulse Secure activity as supporting Chinese government priorities and reported theft of credentials, email and dual-use intellectual property.
  • Preferred entry: internet-facing remote access and delivery appliances. These devices hold session state, user credentials and SAML trust, sit outside most endpoint detection and are often patched slowly.

The practical implication for defenders: if you run SSL VPN, remote access gateways or ADCs in a defence, aerospace or telecoms supply chain, assume these appliances are on APT5's list and plan monitoring accordingly.

Tactics and techniques (MITRE ATT&CK)

The mapping below uses ATT&CK v20 group G1023 and its procedure descriptions. All IDs were checked against the v20 dataset.

TacticTechniqueWhat APT5 did (per ATT&CK and cited reporting)
Resource developmentT1583.005 BotnetAcquired an operational relay box (ORB) network of compromised systems for follow-on activity.
Initial accessT1190 Exploit Public-Facing ApplicationExploited Pulse Secure VPNs and Citrix ADCs, including CVE-2021-22893 and older Pulse Secure flaws such as CVE-2019-11510.
PersistenceT1505.003 Web Shell; T1554 Compromise Host Software BinaryInstalled web shells on VPN appliances and modified the legitimate DSUpgrade.pm upgrade script to reinstall the ATRIUM web shell.
PersistenceT1053.003 Cron; T1136.001 Local Account; T1098.007 Additional Local or Domain GroupsModified crontab entries including in /var/cron/tabs/; created local accounts with Local Administrator rights to keep access through short-cycle credential rotation.
Credential accessT1003.001 LSASS Memory; T1003.002 Security Account Manager; T1056.001 KeyloggingDumped LSASS through Task Manager and with Mimikatz, copied the SAM hive, and used malware with keylogging capability.
Defense impairmentT1685 Disable or Modify ToolsUsed CLEANPULSE to patch a process in memory and prevent certain log events from being written.
StealthT1070 Indicator Removal; T1070.003 Clear Command History; T1036.005 Match Legitimate Resource Name or LocationCleared SSL VPN logs with THINBLOOD, cleared command history on ESXi hosts, named exfiltration archives like Windows updates (KB<digits>.zip).
DiscoveryT1654 Log EnumerationUsed BLOODMINE to parse Pulse Connect Secure logs for logins, message IDs and web requests.
Lateral movementT1021.001 Remote Desktop Protocol; T1021.004 SSHMoved with RDP and SSH, including temporarily enabling SSH on ESXi hosts.
Stealth, initial accessT1078.002 Domain Accounts; T1078.004 Cloud AccountsUsed legitimate domain credentials laterally and stolen credentials to reach Microsoft 365.
CollectionT1074.001 Local Data StagingStaged data, often in C:\Users\Public, before exfiltration as JAR or ZIP archives.

Software: ATT&CK associates APT5 with appliance malware SLOWPULSE (S1104), PACEMAKER (S1109), PULSECHECK (S1108), SLIGHTPULSE (S1110) and RAPIDPULSE (S1113), and with Windows tooling including Mimikatz, Skeleton Key (S0007), PoisonIvy, gh0st RAT and PcShare.

How they abuse identities and credentials

APT5 is an identity problem dressed up as an appliance problem. The VPN or ADC is where user credentials, MFA results and SAML assertions are processed, so whoever controls it controls authentication.

Bypassing authentication on the appliance

Mandiant's analysis of SLOWPULSE described a trojanised shared library on Pulse Secure appliances with four variants. When a hardcoded backdoor password was supplied, the code forced a successful result for LDAP, RADIUS-2FA, ACE-2FA or RealmSignin authentication, giving VPN access without a valid password or second factor. The NSA Citrix advisory similarly warns that the actor can facilitate illegitimate access by bypassing normal authentication controls, and the related Citrix flaw CVE-2022-27518 is described in the CISA KEV catalogue as an authentication bypass affecting ADC and Gateway configured as a SAML service provider or identity provider.

Harvesting credentials in transit

  • On the VPN: SLOWPULSE variant 2 and RADIALPULSE logged plaintext usernames and passwords to files; PACEMAKER read credentials from process memory; BLOODBANK read LMDB files for plaintext passwords during single sign-on (Mandiant).
  • On Windows: Mimikatz was run from the RDP source system over a mapped drive rather than copied to the target, alongside copies of the SAM, SECURITY and SYSTEM hives and LSASS dumps through Task Manager (Mandiant, May 2021).
  • Privileged pairs: Mandiant observed the actors targeting privileged accounts belonging to users whose standard accounts were already compromised. Separating admin and user identities only helps if the admin account cannot be reached from the user's credential.

Persisting with accounts and the cloud

ATT&CK records local administrator accounts created to outlast short-cycle credential rotation, and stolen credentials used against Microsoft 365. The NSA advisory asks defenders to flag legitimate user activity that has no matching SAML token issued by the identity provider, which is the signature of an appliance minting or accepting sessions on its own. For hardening of the federation trust itself see the SAML Security Guide.

Service accounts matter here: Ivanti's guidance, quoted by Mandiant, was to reset all passwords in the environment and review service accounts that could authenticate to the VPN. LDAP bind accounts and RADIUS shared secrets configured on the appliance should be treated as exposed after any compromise. See the Service Account Security Guide.

Notable incidents

WhenEventSource
August 2020 to March 2021UNC2630 compromises Pulse Secure appliances at US Defense Industrial Base companies; 12 malware families documented, later 16.Mandiant, April and May 2021
April 2021CVE-2021-22893, a Pulse Connect Secure use-after-free enabling unauthenticated remote code execution, is disclosed as exploited; Pulse Secure releases the Integrity Checker Tool update (KB44755). CISA later adds it to KEV (3 November 2021).Mandiant; CISA KEV
17 to 20 April 2021Mandiant sees UNC2630 return to dozens of compromised devices and remove web shells such as ATRIUM and SLIGHTPULSE before public disclosure.Mandiant, May 2021
13 December 2022NSA publishes "APT5: Citrix ADC Threat Hunting Guidance". Citrix releases a critical security update the same day; CISA adds CVE-2022-27518 to KEV on 13 December 2022.NSA; CISA KEV

The same appliance classes have been central to other incidents on this site, such as the Ivanti Connect Secure Exploitation 2024 analysis. That case is not attributed to APT5 here, but it shows the same pattern of VPN appliances exposing credentials and service accounts.

Detection and defence

Appliance integrity (highest value)

  • Hash the binaries: for Citrix ADC, NSA lists nsppe, nsaaad, nsconf, nsreadfile and nsconmsg under /netscaler; compare MD5 hashes with vendor or known-good values and investigate any deviation.
  • Check memory mappings: NSA flags an nsppe memory mapping at 0x10400000 with rwx permissions as a sign of tampering.
  • Run the vendor integrity tool: for Pulse Secure and Ivanti Connect Secure, run the Integrity Checker Tool regularly, confirm it supports your version, and manually inspect the rollback partition if the tool was not run before an upgrade. See the Ivanti Connect Secure Hardening and Integrity Checking Guide and the NetScaler ADC and Gateway Hardening Guide.
  • Upgrade from the console: Mandiant recommended upgrading from the appliance console rather than the web interface so that a tampered DSUpgrade.pm is not executed.
  • Keep forensic history: store scheduled tech support bundles or snapshots in offline or immutable storage (NSA).

Log correlation

SignalWhereWhy it matters
VPN login success where LDAP or RADIUS shows failureVPN logs vs directory and RADIUS logsMandiant's key SLOWPULSE indicator.
User session with no matching IdP SAML tokenADC logs vs identity provider sign-in logsNSA indicator of authentication bypass.
Paired "Changing pitboss policy" entries switching X to Y and backCitrix ns.logNSA notes the tooling runs pb_policy twice.
Gaps in logs, or device logs that differ from the remote storeAppliance vs SIEMLog wiping (THINBLOOD) and suppression (CLEANPULSE).
Unexpected crontab changes or files in /var/cron/tabs/Appliance filesystemPersistence (T1053.003).
Files named res* with no or a three-letter extension, often in /tmpCitrix ADC filesystemNSA artefact found in some affected organisations.
Unusual POSTs to meeting_testjs.cgi, or X-KEY, X-CMD, X-CNT headersPulse web logs or proxyPULSECHECK web shell (Mandiant).
SSH enabled on ESXi then disabled, shell history clearedESXi logsLateral movement and T1070.003. See the ESXi Logs Detection Guide.

NSA also published three YARA rules (tricklancer_a, tricklancer_b and tricklancer_c) in the advisory; load them into your appliance and file scanning workflow. Forward dmesg and ns.log off the device as NSA recommends. Routing guidance for appliance syslog is in the Network Device Syslog and Configuration Change Detection Guide.

Architecture and response

  1. Place ADC and VPN management and, where feasible, user portals behind a control that requires valid, ideally multi-factor, authentication, and isolate appliances from the wider environment (NSA).
  2. Patch to current versions even when no compromise is found; track CVE-2022-27518, CVE-2021-22893 and CVE-2019-11510 status from the KEV catalogue.
  3. On any positive result, follow the Edge Device Compromise Investigation Playbook: capture memory and an image first, then restore to a known-good state rather than cleaning in place.
  4. Reset every password that passed through the appliance, rotate LDAP bind credentials and RADIUS secrets, and revoke Microsoft 365 sessions and refresh tokens for affected users.
  5. Hunt for new local administrator accounts and additions to local groups (Windows Event ID 4720 and 4732) on servers reachable from the VPN pool.

Framework mapping: appliance integrity checks map to NIST SP 800-53 SI-7 (Software, Firmware, and Information Integrity); off-device logging to AU-9 (Protection of Audit Information); correlation and hunting to SI-4 (System Monitoring) and CSF 2.0 DE.CM-09; patching edge devices to RA-5; and MFA on remote access to IA-2 and AC-17.

Metrics: percentage of edge appliances with a passing integrity check in the last 30 days; median days from KEV listing to patch on edge devices; percentage of appliance logs forwarded off-box with no gaps over 15 minutes; count of VPN sessions without a matching IdP or directory authentication event (target zero, investigated within 24 hours).

Practitioner checklist

  • Inventory every SSL VPN, remote access gateway and ADC, with version and exposure, and confirm the owner.
  • Schedule vendor integrity checks and binary hash comparison, and store results and tech support bundles immutably.
  • Forward appliance logs, dmesg and ns.log to the SIEM and alert on gaps and on device versus SIEM mismatches.
  • Build the correlation rule: VPN or ADC session without a matching directory, RADIUS or IdP authentication event.
  • Load the NSA tricklancer YARA rules and the Mandiant Pulse Secure rules into your scanning workflow.
  • Patch KEV-listed edge vulnerabilities within days, and upgrade Pulse or Ivanti appliances from the console.
  • Pre-agree a response runbook: image, restore to known-good, reset all credentials and service account secrets that touched the appliance.
  • Hunt for new local admin accounts, LSASS access and SAM hive copies on hosts reachable from VPN address pools.

Standards and references

Related NHI Mgmt Group resources: Threat Actor Groups Guide · Edge Device Security Guide · China State Cyber Operations Guide · Salt Typhoon Threat Actor Profile

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.