APT40 turns newly published proof-of-concept exploits into working intrusions within hours or days, then lives off the credentials it finds on the first server it lands on. In July 2024 Australia's ASD's ACSC led eleven partner agencies in an advisory that assesses the group conducts operations for the People's Republic of China Ministry of State Security (MSS), and that described two Australian intrusions in detail. This profile is for vulnerability management leads, SOC analysts and identity teams in government, defence, research, maritime and critical infrastructure organisations. It covers names and attribution as stated by those agencies and the US Department of Justice, targeting, ATT&CK techniques, the credential abuse pattern, notable incidents and a detection and hardening plan. Broader context on Chinese state activity is in the China State Cyber Operations Guide.
Key takeaways
- Patch speed is the main control. The 2024 advisory says the group weaponises new public exploits within hours or days and recommends patching or mitigating internet-facing systems within 48 hours.
- Public-facing systems over phishing. The agencies state APT40 prefers exploiting vulnerable public infrastructure to techniques that need user interaction, although earlier campaigns relied heavily on spearphishing.
- Credentials are the objective inside. In both 2022 case studies the actor went after service account credentials, Kerberos tickets, cleartext passwords captured at a login portal, MFA token values and session tokens.
- Infrastructure hides in plain sight. Compromised small office and home office (SOHO) routers and legitimate websites act as last-hop redirectors, so IP reputation alone misses the traffic.
- Flat networks turn a DMZ foothold into a breach. One victim's network was flat and an internal application allowed file upload; segmentation and jump servers for Active Directory administration were key recommendations.
Who they are (names and aliases)
MITRE ATT&CK tracks the group as Leviathan (G0065), a Chinese state-sponsored espionage group active since at least 2009, which ATT&CK attributes to the MSS Hainan State Security Department and an affiliated front company. The July 2024 joint advisory states that the authoring agencies assess the group conducts malicious cyber operations for the MSS and that it was previously reported as based in Haikou, Hainan Province, receiving tasking from the Hainan State Security Department. In July 2021 the US Department of Justice unsealed an indictment alleging that three Hainan State Security Department officers and one hacker ran a campaign from 2011 to 2018 through a front company, Hainan Xiandun Technology Development Co., Ltd.; the department notes the charges are allegations.
| Name | Used by | Notes |
|---|---|---|
| APT40 | Mandiant / Google, joint advisory | Most common public name |
| Leviathan | Proofpoint, MITRE ATT&CK | ATT&CK primary name (G0065) |
| Gingham Typhoon (earlier Gadolinium) | Microsoft | Listed as overlapping in the 2024 advisory |
| Kryptonite Panda | CrowdStrike | Listed as overlapping in the 2024 advisory |
| Bronze Mohawk | Secureworks | Listed as overlapping in the 2024 advisory |
| TEMP.Periscope, TEMP.Jumper, MUDCARP | Earlier vendor names | Associated names in ATT&CK |
Vendor names do not always map one to one: the advisory itself says the activity "overlaps" with these groups. Treat alias lists as a starting point for intelligence correlation, not proof that two reports describe the same operators.
Targeting and motivation
- Sectors: ATT&CK lists academia, aerospace and aviation, biomedical, defence industrial base, government, healthcare, manufacturing, maritime and transportation. The 2021 indictment adds infectious disease research (Ebola, MERS, HIV/AIDS, Marburg and tularemia).
- Geography: the indictment names twelve countries including the United States, United Kingdom, Canada, Germany, Norway, Switzerland, Malaysia, Indonesia, Cambodia and Saudi Arabia. The 2024 advisory is written from repeated Australian targeting, with ASD's ACSC stating the group continues to target Australian networks.
- Motivation: state intelligence collection, including commercial and research data with economic value, according to the indictment.
- Selection method: continuous reconnaissance for vulnerable, end-of-life or unmaintained devices on networks of interest (T1595.002 Vulnerability Scanning), then exploitation as soon as a usable exploit exists. Exposure, not organisation size, decides who is hit.
Tactics and techniques (MITRE ATT&CK)
Techniques mapped to G0065 in ATT&CK v20 are marked (G). Techniques marked (A) are listed by the 2024 joint advisory for its case studies.
| Tactic | Technique | How it shows up |
|---|---|---|
| Reconnaissance | T1595.002 Vulnerability Scanning (G); T1594 Search Victim-Owned Websites (A) | Ongoing scanning for unpatched and end-of-life devices |
| Resource development | T1584.008 Network Devices (G) | Compromised SOHO routers used as operational infrastructure and last-hop redirectors |
| Initial access | T1190 Exploit Public-Facing Application (G) | Log4j CVE-2021-44228, Confluence CVE-2021-26084, Exchange CVE-2021-31207, CVE-2021-34523 and CVE-2021-34473; a custom web application; a remote access login portal |
| Initial access | T1133 External Remote Services; T1078 Valid Accounts (G) | VPN access and valid credentials |
| Initial access (historic) | T1566.001 Spearphishing Attachment; T1189 Drive-by Compromise (G) | Office documents and watering holes in the 2011 to 2018 period |
| Persistence | T1505.003 Web Shell (G) | Web shells for early persistence, including Java memory shells uploaded by ASD to VirusTotal |
| Credential access | T1552.001 Credentials In Files; T1558.003 Kerberoasting (A) | Hardcoded service account password; password files; Kerberoasting after AD enumeration |
| Credential access | T1056.003 Web Portal Capture; T1111 Multi-Factor Authentication Interception; T1040 Network Sniffing (A) | Modified authentication process captured hundreds of cleartext credential pairs and MFA token values; JWTs captured with tcpdump |
| Lateral movement | T1021.002 SMB/Windows Admin Shares; T1563.002 RDP Hijacking (A); T1021.001 RDP (G) | SMB shares mounted from DMZ hosts; captured tokens could hijack VDI sessions |
| Command and control | T1572 Protocol Tunneling; T1090.003 Multi-hop Proxy (G); T1001.003 Protocol or Service Impersonation (A) | Secure Socket Funnelling tool, multi-hop proxies, Tor |
| Stealth, exfiltration | T1027.003 Steganography; T1567.002 Exfiltration to Cloud Storage; T1102.003 One-Way Communication (G) | Stolen data hidden on GitHub; Dropbox uploads with API keys; C2 instructions from GitHub and TechNet profiles |
How they abuse identities and credentials
The two Australian case studies in the 2024 advisory are an identity story as much as a patching story. The exploit gets the actor onto one host; credentials get it everywhere else.
| Observed behaviour | Case study | Identity control that breaks it |
|---|---|---|
| Tested a service account credential likely hardcoded in internal binaries | 1 (July 2022) | No secrets in binaries or config; use managed service accounts (the advisory recommends them) or a vault; scan builds for embedded credentials |
| Logged in with a compromised user account from the DMZ | 1 | DMZ hosts cannot authenticate to the internal directory with user accounts; MFA for remote access |
| Kerberoasting to obtain network credentials | 1 | Long random passwords or gMSA for every account with an SPN; AES-only encryption types; alert on RC4 ticket requests |
| Exfiltrated password files and privileged credentials | 1 | Remove credential spreadsheets from file shares; rotate every exposed secret after eviction |
| Altered the login process on a remote access appliance to capture cleartext credentials and MFA token values | 2 (April to May 2022) | Phishing-resistant MFA that does not reveal a reusable value; file integrity monitoring on appliances; force a full credential reset for all portal users after compromise |
| Captured JWTs with tcpdump, usable to hijack VDI sessions | 2 | Short token lifetimes, sender-constrained tokens where supported, session revocation as part of the response |
| SSH brute force from the compromised segment | 2 | Key-only SSH, no shared root passwords, rate limits; see the OpenSSH Server Hardening Guide |
- Service accounts are the bridge: a hardcoded credential with network rights is worth more to this actor than an end user's mailbox. See the Service Account Security Guide and the Group Managed Service Accounts (gMSA) Guide.
- Kerberoasting follows enumeration: the actor enumerated Active Directory before requesting service tickets. See the Kerberoasting and AS-REP Roasting Defence Guide.
- The appliance is an identity provider: a remote access gateway that handles logins sees every password and MFA code. Treat it as Tier 0, with integrity monitoring and fast patching.
Notable incidents
| Date | Activity | Source |
|---|---|---|
| 2011 to 2018 | Campaign against aviation, defence, education, government, health, biopharmaceutical and maritime targets in twelve countries; data hidden on GitHub with steganography; Dropbox API keys used to exfiltrate | US DOJ indictment, unsealed July 2021 |
| July 2021 | Public attribution of MSS-linked activity by the United States and partners; CISA advisory AA21-200A on APT40 tradecraft | US DOJ; cited in ATT&CK G0065 |
| April to May 2022 | Australian organisation: remote access appliance hosts compromised, credentials, MFA values and JWTs captured | ASD's ACSC joint advisory, case study 2 |
| July to September 2022 | Australian organisation: custom web application exploited, Kerberoasting, building management documentation and privileged credentials exfiltrated | ASD's ACSC joint advisory, case study 1 |
| 9 July 2024 | Joint advisory "APT40 Advisory: PRC MSS tradecraft in action" from eight countries' agencies | ASD's ACSC and partners |
Detection and defence
Detection signals
| Signal | Where | Notes |
|---|---|---|
Process execution from C:\Users\Public\, C:\Windows\Temp\ and other world-writable Windows folders | EDR or Sysmon process creation | The advisory publishes three experimental Sigma rules for these paths; expect false positives |
| New or modified JSP, ASPX or PHP files and Java memory shells on web servers | File integrity monitoring, web server logs | Web shells are the usual first persistence |
| Modified login scripts or binaries on remote access appliances; unexpected iptables rules or listening ports | Vendor integrity checkers, appliance logs | Case study 2 opened ports 9998 and 9999 |
| Bursts of Kerberos service ticket requests (Event ID 4769), especially RC4 | Domain controller security logs | Kerberoasting after AD enumeration |
| DMZ hosts authenticating to internal shares or domain controllers | Firewall and Windows logon events | Should be close to zero by design |
| Long-lived outbound tunnels from servers to residential or SOHO address space | Proxy, NetFlow, firewall | SSF tunnel reconnected for weeks in case study 1 |
Hardening plan
- Now: inventory every internet-facing system and its owner, and confirm CVE-2021-44228, CVE-2021-26084 and the Exchange ProxyShell set are fixed; all are in the CISA KEV catalog. Replace end-of-life devices that cannot be patched.
- Within a month: set a 48-hour patch or mitigate target for internet-facing systems, matching the advisory, with an emergency change path tested in a pilot group first.
- Within a quarter: segment the DMZ from internal networks, administer Active Directory only from monitored jump servers, and move service accounts with SPNs to gMSA or long random passwords.
- Ongoing: centralise and retain web server, Windows event and proxy logs, deploy the advisory's Sigma rules in alert-only mode, tune for two weeks, then promote to high-priority alerts.
Rollback: emergency patches can break appliances; take configuration backups and know the vendor's rollback image before patching, but do not leave a known-exploited version exposed while deciding. Apply vendor mitigations or take the service offline instead.
Metrics
- Median time to patch KEV-listed CVEs on internet-facing systems: target 48 hours or less.
- End-of-life internet-facing devices: target zero.
- Service accounts with SPNs and human-set passwords: trending to zero.
- Hardcoded credentials found per build scan: trending to zero.
- Web shell detection coverage: percentage of internet-facing web servers with file integrity monitoring.
Framework mapping
| Framework | Reference | Relevance |
|---|---|---|
| NIST SP 800-53 Rev. 5 | SI-2 Flaw Remediation, RA-5 Vulnerability Monitoring and Scanning | Fast patching of exposed services |
| NIST SP 800-53 Rev. 5 | IA-5 Authenticator Management, AC-6 Least Privilege, SI-4 System Monitoring | Service account secrets, segmentation of rights, detection |
| NIST CSF 2.0 | ID.RA-01, PR.PS-02, DE.CM-01 | Vulnerability identification, software maintenance and replacement, network monitoring |
| MITRE D3FEND | D3-SFA System File Analysis | Finding web shells and modified appliance files |
For the response once a web shell is found, use the Web Shell Response Playbook and the Edge Device Compromise Investigation Playbook.
Practitioner checklist
- Subscribe to ASD's ACSC, CISA and NCSC advisories and the CISA KEV feed, and route new entries to the owners of matching systems the same day.
- Keep an up-to-date list of internet-facing systems with owner, version and end-of-life date.
- Set and measure a 48-hour patch or mitigate target for internet-facing services.
- Deploy file integrity monitoring on web servers and remote access appliances, and run vendor integrity tools after every advisory.
- Scan code, binaries and configuration for hardcoded credentials and move them to managed service accounts or a vault.
- Block DMZ hosts from authenticating to internal directory services and file shares.
- Alert on RC4 Kerberos service ticket requests and remove RC4 where possible.
- After any appliance compromise, reset every credential that passed through it and revoke active sessions.
Standards and references
- ASD's ACSC and partners: APT40 Advisory, PRC MSS tradecraft in action
- MITRE ATT&CK: Leviathan (G0065)
- US Department of Justice: Four Chinese nationals working with the Ministry of State Security charged with global computer intrusion campaign
- MITRE ATT&CK: Exploit Public-Facing Application
- MITRE ATT&CK: Steal or Forge Kerberos Tickets: Kerberoasting
- CISA: Known Exploited Vulnerabilities Catalog
- NIST: SP 800-53 Rev. 5
Related NHI Mgmt Group resources: Threat Actor Groups Guide · Edge Device Exploitation by Threat Actors Guide · Volt Typhoon Threat Actor Profile · Salt Typhoon Threat Actor Profile