Join our Newsletter — 33% off our NHI Course
Home› Guides› ToddyCat Threat Actor Profile: Exchange Intrusions, Domain Admin…
Guide Threat Intelligence & Threat Actors

ToddyCat Threat Actor Profile: Exchange Intrusions, Domain Admin Abuse and Defences

PDF for newsletter subscribers · free to read online

← All guides
By NHI Mgmt Group, edited by Lalit Choda Updated 20 November 2026 15 min read
On this page

ToddyCat is an espionage cluster, tracked by MITRE ATT&CK as G1022, that broke into Microsoft Exchange servers from late 2020 and then used stolen domain admin credentials, scheduled tasks and cloud storage tokens to collect documents from government and military targets across Asia and Europe. This profile sets out what Kaspersky's primary research and the ATT&CK group entry say about the group, how it abuses identities and credentials, and what security and identity teams can do to detect and contain it.

Everything here comes from two sources: Kaspersky's original ToddyCat reports on Securelist (June 2022 and October 2023) and the MITRE ATT&CK entry for ToddyCat (G1022). No government advisory attributes ToddyCat to a state, and Kaspersky itself states that it could not attribute the activity to a known group. This page does not go further than those sources, and it names no individuals. For the wider context of how NHI Mgmt Group profiles threat actors, see the Threat Actor Groups Guide.

Key takeaways

  • ToddyCat first appeared in Kaspersky telemetry on 22 December 2020, compromising Exchange servers with an exploit that led to a China Chopper web shell, then scaled up sharply from late February 2021 using ProxyLogon (CVE-2021-26855).
  • Kaspersky describes the targets as mostly government organisations, military entities and military contractors, with a strong interest in Southeast Asia and activity in other parts of Asia and Europe.
  • After initial access, the group's main lever is identity: it enumerates domain admins, mounts remote administrative shares with compromised domain admin credentials, and, according to Kaspersky, rotates the credentials it uses over time.
  • Remote execution relies on scheduled tasks running as SYSTEM that are created, run and deleted along with the network share, which makes short lived task and share events the best detection opportunity.
  • Exfiltration uses cloud storage: a DropBox uploader and a tool Kaspersky calls Pcexter that uploads archives to OneDrive using OAuth refresh tokens, a clear non-human identity (NHI) concern.
  • Kaspersky could not attribute ToddyCat to a known group and explicitly declined to merge it with an overlapping cluster, so defenders should plan around behaviours rather than a nation-state label.

Who they are (names and aliases)

ToddyCat is the name Kaspersky gave to the cluster in its June 2022 Securelist report, and it is the only name ATT&CK lists for group G1022. ATT&CK describes ToddyCat as a sophisticated threat group active since at least 2020 that uses custom loaders and malware in multi-stage infection chains against government and military targets across Europe and Asia.

Kaspersky's first report notes that ESET tracked the ProxyLogon related cluster as "Websiic", and that GTSC published a report on the infection vector. Those are related observations from other researchers, not formal aliases in the ATT&CK entry, so this profile treats them as context only.

On attribution, Kaspersky's wording is direct: it was "not able to attribute the attacks to a known APT group". It observed that the victims resemble those of Chinese-speaking groups, and that three organisations were compromised in a similar timeframe by both ToddyCat and a group using the FunnyDream backdoor. Kaspersky then states that, despite the overlap, it does not feel confident merging ToddyCat with the FunnyDream cluster. This profile follows that position: ToddyCat is an unattributed espionage cluster.

Targeting and motivation

Kaspersky's June 2022 report describes the targets as mostly government organisations, military entities and military contractors. Its timeline shows how the targeting widened:

  • December 2020 to February 2021: a small set of servers at three organisations in Taiwan and Vietnam.
  • Late February to early March 2021: after ProxyLogon became available, organisations in Afghanistan, India, Iran, Malaysia, Pakistan, Russia, Slovakia, Thailand and the United Kingdom.
  • After May 2021: Kyrgyzstan, Uzbekistan and Indonesia, alongside most of the Asian countries above.
  • From September 2021: loaders delivered through Telegram to desktop systems in Central Asia.

Kaspersky's October 2023 report describes ToddyCat as responsible for multiple sets of attacks against high-profile entities in Europe and Asia. On motivation, Kaspersky assesses that the activity is "probably used to achieve critical goals, likely related to geopolitical interests". That is a probable assessment from the researcher, not a confirmed motive, and the toolset (document collectors, archive tools and cloud uploaders) is consistent with intelligence collection rather than financial gain.

For security and identity teams, the practical point is that the group goes after organisations that hold sensitive government, defence or diplomatic information and that run internet-facing Exchange servers or similar collaboration infrastructure.

Tactics and techniques (MITRE ATT&CK)

The table below lists techniques that ATT&CK v20 maps to ToddyCat, grouped by tactic, with the behaviour each mapping describes. IDs were checked against the local ATT&CK v20 cache.

TacticTechniqueObserved behaviour (per ATT&CK and Kaspersky)
Initial accessT1190 Exploit Public-Facing ApplicationExploited ProxyLogon (CVE-2021-26855) to compromise Exchange servers at multiple organisations.
Initial accessT1566.003 Spearphishing via ServiceSent loaders configured to run Ninja as zip archives via Telegram.
ExecutionT1059.001 PowerShell, T1059.003 Windows Command ShellPowerShell scripts for collection; .bat scripts and cmd for execution.
ExecutionT1047 Windows Management InstrumentationWMI used to run scripts for document collection.
Execution and persistenceT1053.005 Scheduled TaskScheduled tasks used to run discovery commands and collection scripts.
ExecutionT1106 Native APIWinExec used to run commands received from command and control.
Defence evasionT1036.005 Match Legitimate Resource Name or LocationMalware components named debug.exe.
Defence evasionT1564.003 Hidden WindowScripts run with powershell.exe -windowstyle hidden.
Defence evasionT1686 Disable or Modify System FirewallAdded an inbound firewall rule named SGAccessInboundRule allowing UDP 49683 before running a backdoor.
Credential useT1078.002 Domain AccountsUsed compromised domain admin credentials to mount network shares.
DiscoveryT1069.002 Domain Groups, T1087.002 Domain AccountRan net group "domain admins" /dom and net user /dom.
DiscoveryT1018, T1049, T1057, T1083, T1518.001, T1680ping, netstat, tasklist, document enumeration, a WMI check for a specific security product and collection of drive details.
Lateral movementT1021.002 SMB/Windows Admin SharesLocally mounted network shares used for lateral movement.
CollectionT1005, T1074.002, T1560.001Document collection, staging on an exfiltration host with xcopy and archiving with 7-Zip and RAR.
Command and controlT1095 Non-Application Layer ProtocolA passive backdoor that receives commands in UDP packets.
ExfiltrationT1567.002 Exfiltration to Cloud StorageA DropBox uploader used to exfiltrate stolen files.

ATT&CK also associates the following software with ToddyCat: China Chopper, Samurai, Ninja, LoFiSe, Pcexter, Cobalt Strike, and the built-in Net, Ping and netstat utilities.

The Exchange infection chain

Kaspersky's 2022 report describes a consistent chain on Exchange servers. A China Chopper web shell downloaded and ran a dropper named debug.exe. The dropper decrypted a file called debug.xml and installed a .NET loader and a DLL loader, then established persistence by forcing the legitimate svchost.exe to load a malicious service named WebUpdate. The DLL loader passed an encrypted payload from the registry to the .NET loader, which decoded it and ran it in memory.

The final payload was Samurai, a modular C# backdoor that receives HTTP POST requests on ports 80 and 443, compiles attacker-supplied C# code at runtime and uses encrypted communications. Because Samurai runs inside svchost.exe and shares ports 80 and 443 with Exchange, Kaspersky notes it is not visible through process listing or port scanning alone. In some cases Samurai deployed Ninja, a C++ post-exploitation agent with process management, reverse shells, code injection, plugin loading, TCP proxying and a local server for pivoting.

The 2023 toolset

Kaspersky's October 2023 report documents a second generation of tooling: standard loaders run through rundll32.exe or side-loaded via a legitimate media player executable; a tailored loader whose decryption key incorporates the system's drive model and volume GUID so the payload only runs on the intended host; a newer Ninja version; LoFiSe, which watches drives for new or changed documents and packs them into password-protected ZIP files every three hours; a DropBox uploader; Pcexter; a passive UDP backdoor whose purpose Kaspersky describes as currently unknown; and Cobalt Strike, observed shortly before Ninja was deployed.

How they abuse identities and credentials

ToddyCat's post-exploitation work is mostly identity work. Kaspersky's 2023 report describes a repeatable pattern that security and identity teams should recognise.

Domain admin enumeration

The group enumerates domain admins, domain computers and domain controllers with built-in tools such as net and ping. ATT&CK records net group "domain admins" /dom and net user %USER% /dom. These are ordinary administrative commands, which is why they work: they blend into normal activity unless the identity running them, the host they run from and the timing are checked together.

Domain admin credentials as the lateral movement engine

Kaspersky states that the group moves laterally by mounting remote administrative shares (c$) with "compromised domain admin credentials", and that "the attackers take care to rotate the credentials used over time". In other words, the group holds more than one privileged credential and switches between them, which defeats any response that resets a single account and assumes the intrusion is over.

The report does not explain how those domain admin credentials were first obtained, and this profile does not guess. The defensive conclusion is the same either way: highly privileged domain accounts that can log on to ordinary servers and mount c$ shares across the estate are the single biggest multiplier for this group. Teams that want to check their own exposure can start with How to find service accounts with Domain Admin rights, because service accounts with standing Domain Admin membership are a common route to the same outcome.

Scheduled tasks as SYSTEM

Remote execution follows a tight loop. Kaspersky describes scheduled tasks that run as SYSTEM on remote hosts, where "a scheduled task is created, executed and immediately deleted along with the network share", repeated for each target. Task names such as "one" and "tpcd" are reused within a session, while script names are short random strings. Command output is written to a file on the remote host so the operator can read it through the mounted share. The loop leaves only a brief trace on each host, but it always involves a privileged logon, a share mount and a task creation and deletion within seconds.

OAuth refresh tokens for exfiltration

Pcexter, according to Kaspersky, waits for a LoFiSe event and uploads archives to OneDrive using OAuth refresh tokens. This is a non-human identity in the attacker's hands: a long lived token for a cloud storage account that lets malware upload data without an interactive sign-in. From the victim's perspective the traffic goes to a legitimate cloud service, so proxy categories alone will not flag it. The How to detect use of a stolen OAuth refresh token guide covers the token side of this problem in more depth.

Share creation on the staging host

Kaspersky also observed the group temporarily creating a share named tmp on the staging host, pointing to c:\windows\debug, and deleting it afterwards. Collected documents were archived with RAR or 7-Zip, often split into 200 MB volumes, before exfiltration through Pcexter, the DropBox uploader or other public storage services.

Notable incidents

Kaspersky's reports do not name individual victim organisations, and this profile does not attempt to identify them. The documented campaigns are:

  • Exchange intrusions, December 2020 onwards: compromise of Exchange servers at three organisations in Taiwan and Vietnam using an exploit Kaspersky did not name, leading to a China Chopper web shell. Kaspersky suspects exploitation began in December 2020 but states it lacks enough information to confirm the vector for that period.
  • ProxyLogon wave, late February to March 2021: the same infection chain deployed through ProxyLogon against organisations across Europe and Asia, followed by Samurai and in some cases Ninja.
  • Telegram delivered loaders, from September 2021: executables inside zip archives sent through Telegram to desktop systems in Central Asia, decrypting a Ninja payload.
  • Post-exploitation campaigns reported in October 2023: domain admin share mounting, SYSTEM scheduled tasks, LoFiSe collection and cloud exfiltration through Pcexter and the DropBox uploader, against high-profile entities in Europe and Asia.

ProxyLogon itself is catalogued as a known exploited vulnerability. Teams running on-premises Exchange should treat the CISA Known Exploited Vulnerabilities Catalog as the minimum patch priority list and follow the Microsoft Exchange Server Security Guide for hardening.

Detection and defence

Kaspersky's reports do not contain a formal detection section, but they list enough artefacts to build hunts. Treat any specific indicator (IP address, domain, file name) as historical unless current threat intelligence confirms it is active; the behaviours are more durable than the indicators.

Exchange and web-facing servers

  • Patch Exchange to a supported, current cumulative update and confirm ProxyLogon class vulnerabilities are closed. Where Exchange cannot be patched quickly, isolate it.
  • Hunt for web shells under Exchange and IIS paths using the approach in How to Hunt for Web Shells on IIS, Apache and Nginx.
  • Kaspersky notes that running netsh http show servicestate verbose=yes may reveal suspicious registered URLs, such as those under /OWA/AUTH/TOKEN/, which helps find a backdoor that shares ports with Exchange.
  • Review services loaded by svchost.exe for unexpected entries; Kaspersky's IoCs include a service named WebUpdate and, in 2023, FontCacheSvc parameters and a SvcHost value of fontcsvc.

Privileged identity activity

  • Alert on domain admin or other Tier 0 accounts logging on to member servers and workstations. If your design keeps Tier 0 accounts on domain controllers and admin workstations only, any other logon is a strong signal. The Active Directory Tiering vs Red Forest guide explains the model.
  • Correlate net use against c$ or admin$ shares with remote scheduled task creation and deletion within a short window on the same target host.
  • Alert on scheduled tasks created remotely to run as SYSTEM that are deleted shortly after running, especially with short generic names.
  • Watch for creation of new SMB shares on servers (for example a share named tmp pointing at a Windows system directory) followed by deletion.
  • Baseline domain group enumeration such as net group "domain admins" /dom and flag it from unexpected hosts or accounts. Active Directory Security Monitoring and Event IDs Guide lists the relevant Windows events.

Host and network artefacts

  • Alert on new Windows firewall rules added from the command line, including a rule named SGAccessInboundRule allowing inbound UDP on port 49683.
  • Flag WMI process queries that look for specific security product processes, a sign of pre-deployment checks.
  • Watch for PowerShell launched with a hidden window from scheduled tasks or WMI.
  • Flag rundll32.exe loading DLLs from unusual directories and legitimate signed executables side-loading DLLs from user writable paths.

Cloud exfiltration and tokens

  • Restrict which cloud storage services servers can reach. Servers that hold sensitive data rarely need direct access to personal or third-party cloud storage.
  • Where corporate cloud storage is allowed, require that uploads originate from managed identities and known tenants, and alert on OAuth token use for storage accounts outside your tenant.
  • Look for large split archives (Kaspersky observed 200 MB volumes) created on servers, followed by outbound transfer.

Response

Because Kaspersky reports that ToddyCat rotates the credentials it uses, a response that resets one domain admin password is not enough. Plan to reset all Tier 0 credentials, including service accounts with domain admin rights, in a coordinated window, and to revoke OAuth tokens and app consents associated with any cloud storage used for exfiltration. The NCSC incident response processes are a useful structure for that work.

Common failure patterns

  • Treating Exchange as patched once: organisations that patched ProxyLogon but never hunted for web shells or malicious services left the intruder in place, because the backdoor persists independently of the vulnerability.
  • Domain admins logging on everywhere: where domain admin accounts are used for routine server administration, their credentials are available on many hosts and every mounted c$ share looks normal.
  • Single account reset: resetting the one account seen in the logs while the attacker holds and rotates several others.
  • Ignoring short lived tasks and shares: monitoring that only looks at persistent scheduled tasks misses tasks created, run and deleted in seconds.
  • Trusting cloud storage traffic: allowing servers to talk to public storage services because they are well-known brands, without checking which identity and tenant are in use.
  • Waiting for an attribution label: deprioritising an unattributed cluster because it lacks a familiar state sponsor name, even though its behaviours are well documented.

Practitioner checklist

  • Confirm all on-premises Exchange servers are on a supported, current update and that known exploited vulnerabilities are closed.
  • Hunt Exchange and IIS hosts for web shells, unexpected svchost.exe services and unusual HTTP.sys URL registrations.
  • Inventory every account and service account in Domain Admins and other Tier 0 groups, and remove standing membership that is not needed.
  • Restrict Tier 0 account logons to domain controllers and dedicated admin workstations, and alert on any other logon.
  • Build a correlation rule for admin share mount, remote scheduled task as SYSTEM and task deletion on the same host within minutes.
  • Alert on command line firewall rule changes and new SMB share creation on servers.
  • Restrict server egress to public cloud storage and monitor OAuth token use for storage outside your tenant.
  • Prepare a coordinated Tier 0 credential reset runbook that assumes the attacker holds several credentials.
  • Map ToddyCat's ATT&CK techniques to your detection coverage and test the gaps.

Standards and references

Related NHI Mgmt Group resources: Threat Actor Groups Guide · Threat Intelligence for Identity and NHI Guide · APT5 Threat Actor Profile Guide · Microsoft Exchange Server 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 20 November 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.