CRAT
- Type
- RAT
- Capabilities
- Credential theftKeyloggingExfiltrationPersistencePost exploitationDefense evasionReconnaissanceProcess injection
- Delivery
- PhishingSpearphishingLure
- Platforms
- Windows
CRAT is a modular Windows remote access trojan first observed in 2020 and used in targeted intrusions against South Korean users and organizations. It has been associated in multiple investigations with Lazarus-linked or broader North Korea-aligned tradecraft, and later reporting connected historical CRAT activity with the Larva-26005 cluster and overlap with Xctdoor and Hansom ransomware operations.
CRAT functions as a standalone RAT and also supports downloading and activating additional plugins from command-and-control infrastructure. Documented capabilities include system reconnaissance, collection of host and security-product information, file and directory enumeration, file read/write and movement operations, command execution, reverse shell access, and data exfiltration. Observed plugin functionality includes screen capture, clipboard monitoring, keylogging, and deployment of the Hansom ransomware component. In some cases, CRAT operators also used credential-stealing tooling to extract browser-stored credentials.
The malware uses multiple defense-evasion and anti-analysis measures, including string obfuscation, dynamic API resolution, runtime code patching, anti-debugging checks, and anti-infection logic based on host artifacts and analysis-tool blocklists. It commonly injects or reflectively loads itself into legitimate Windows processes and can activate DLL plugins through native Windows utilities or in-memory loading. Persistence has been established through scheduled tasks, services, shortcut files, and autorun mechanisms, often via dedicated EXE or DLL loaders that ultimately load the CRAT payload.
Observed delivery has included spear-phishing with malicious Hangul Word Processor documents exploiting CVE-2017-8291, as well as trojanized software distributed through South Korean community sites. Lures have included Korean-language themes such as public-health and business-related subjects. Reporting on later activity indicates operational continuity between CRAT-era campaigns and subsequent Xctdoor intrusions, including shared obfuscation patterns, similar installation conventions, and overlap with Hansom ransomware deployment. CRAT is therefore best understood as a modular RAT platform used in targeted espionage and financially motivated operations against Korean-speaking victims.
Associated vulnerabilities
1 CVE Mallory has correlated with this family across public research and vendor advisories. Each card links to the full Mallory page for that vulnerability.
Associated threat actors
2 distinct threat actors Mallory has associated with this family across public reporting. Open in Mallory to see the full evidence chain and overlapping campaigns.
Cisco Talos has observed a new version of a remote access trojan (RAT) family known as CRAT. Apart from the prebuilt RAT capabilities, the malware can download and deploy additional malicious plugins on the infected endpoint.
"Analysis of the Connection Between Xctdoor and Past CRAT Attack Cases (Larva-26005)" published by Ahnlab. #CVE20178291, #Phishing, #LNK, #Xctdoor, #CRAT, #Larva26005, #Hansom
Techniques & procedures
41 distinct techniques documented for this family, organized by ATT&CK tactic.
Initial Access
2 techniques
Initial Access
"Analysis of the Connection Between Xctdoor and Past CRAT Attack Cases (Larva-26005)" published by Ahnlab. #CVE20178291, #Phishing, #LNK, #Xctdoor, #CRAT, #Larva26005, #Hansom
The first version of CRAT has been known to be distributed via malicious HWPs. The HWPs masquerade as a COVID-19 themed document pertaining to an infectious disease management support group from South Korea.
Execution
8 techniques
Execution
Establish persistence for itself using a scheduled task on the infected endpoint. Usually scheduled to run once every few minutes.
Execute commands with two variations of the functionality: Simply execute the command on the endpoint. Execute a command and send output of command to C2.
Shellcode executing PowerShell to download and activate CRATv1 via regsvr32.
Enumerate drives using the 'dir' command: cmd /c "dir <drive_name> /s >> %temp%\<custom_prefix>error.log"
This variant will create a VBS file to run rar.exe to extract CRAT components.
The HWPs consisted of an exploit for CVE-2017-8291 used to activate malicious shellcode.
"Analysis of the Connection Between Xctdoor and Past CRAT Attack Cases (Larva-26005)" published by Ahnlab. #CVE20178291, #Phishing, #LNK, #Xctdoor, #CRAT, #Larva26005, #Hansom
It attempts to connect to a named pipe named “\\.\Pipe\ChromeUpdatePipe” and, if successful, transmits the payload; the co-installed injector module can read from that named pipe and inject the received payload into a legitimate process.
Persistence
4 techniques
Persistence
Establish persistence for itself using a scheduled task on the infected endpoint. Usually scheduled to run once every few minutes.
Create and run a malicious service on the endpoint to run itself.
The loader may also obtain persistence using a registry RUN key.
여기까지의 과정이 끝나면 지속성 유지를 위해 시작 프로그램 경로에 바로가기를 생성하고 실행하는 데 XcLoader를 RegSvr32를 통해 실행시키는 명령이다.
Privilege Escalation
5 techniques
Privilege Escalation
Establish persistence for itself using a scheduled task on the infected endpoint. Usually scheduled to run once every few minutes.
Depending on the MS Windows OS version, it will inject and reflectively-load itself into the right process.
Create and run a malicious service on the endpoint to run itself.
The loader may also obtain persistence using a registry RUN key.
여기까지의 과정이 끝나면 지속성 유지를 위해 시작 프로그램 경로에 바로가기를 생성하고 실행하는 데 XcLoader를 RegSvr32를 통해 실행시키는 명령이다.
Stealth
12 techniques
Stealth
The RAT consists of multiple obfuscation techniques to hide strings, API names, command and control (C2) URLs and instrumental functions, along with static detection evasion.
Runtime Code Patching: The malware consists of a select few instrumental subroutines that are decoded, executed and then patched/encoded again (during runtime).
API Resolution: All APIs used are resolved dynamically. The process of selecting the API to call is also a cumbersome algorithm... involving index tables, switch cases and API name deobfuscation.
The attackers have also used file names and export API names in the CRAT DLLs to masquerade the RAT as a benign application's library.
Depending on the MS Windows OS version, it will inject and reflectively-load itself into the right process.
This command also deletes the target file once it has been added to the archive.
The implant will also timestomp the tmp file to either a hardcoded value or copy the filetimes from a system file such as %windir%\system32\user32.dll.
For DLL plugins: Activate using regsvr32.exe
For DLL plugins: Activate using rundll32.exe
The implant performs checks to verify its execution on an allowed endpoint. The implant uses a variety of criteria to identify a blocklisted system by utilizing: Process name blocklists Network adapter name blocklists MAC Address blocklists Analysis tool names blocklists
Depending on the MS Windows OS version, it will inject and reflectively-load itself into the right process.
The implant also checks for indicators that the process is being debugged (including CheckRemoteDebuggerPresent). If the infected endpoint fails any of the anti-infection checks, the implant quits execution.
Credential Access
3 techniques
Credential Access
The keylogger plugin monitors the state of all the alphanumeric keys pressed.
STEAL BROWSER PASSWORDS: CHROMEPASS Execute chromepass.exe with the /stext switch to obtain the usernames and passwords stored in Google Chrome.
Execute chromepass.exe with the /stext switch to obtain the usernames and passwords stored in Google Chrome.
Discovery
6 techniques
Discovery
TCP/IP enabled MAC Addresses using WMI query: wmic PATH Win32_NetworkAdapterConfiguration WHERE IPEnabled=TRUE GET MACAddress.
Flag if the current user has administrative privileges
Collects the following system information: Installed AntiVirus software names Installed FirewallProduct names Domain Names... Path to the system folder Flag if the current user has administrative privileges
Collects size information about all files and folders on the infected endpoint... Enumerate drives using the 'dir' command | CRAT has the capability to read the contents of the file specified by the C2 and exfiltrate these to the C2.
The implant performs checks to verify its execution on an allowed endpoint. The implant uses a variety of criteria to identify a blocklisted system by utilizing: Process name blocklists Network adapter name blocklists MAC Address blocklists Analysis tool names blocklists
The implant also checks for indicators that the process is being debugged (including CheckRemoteDebuggerPresent). If the infected endpoint fails any of the anti-infection checks, the implant quits execution.
Collection
4 techniques
Collection
The keylogger plugin monitors the state of all the alphanumeric keys pressed.
The screen capture plugin will capture the current foreground window (every second or so) and save the screenshot to a .tmp file in a hardcoded directory.
This plugin will read the clipboard data and write the contents of the clipboard to a log file.
The log file is then rolled up into a '.rar' archive using a pre-existing installation of a RAR archiver program... The RAR archive is then read to memory and exfiltrated to the C2
Command and Control
4 techniques
Command and Control
Both versions of CRAT use HTTP to communicate with the C2 servers.
Downloads and installs a malicious plugin from the C2.
Generate a random four-byte value (DWORD). XOR the data to be exfiltrated with the DWORD. Base64-encode... Exfiltrate the encoded data as URL encoded form data.
REVERSE SHELL Open up a reverse command shell for arbitrary command execution using Windows pipes.
Exfiltration
1 technique
Exfiltration
When transmitting data, it is segmented... and sent to the C2 server... 3 Transmission of shell command results... 17 Transmit keylog data... 18 Send data whenever the clipboard content changes... 24 Submitting Screenshots
Impact
1 technique
Impact
Cisco Talos의 보고서에 따르면 Hansom 랜섬웨어를 설치해 감염 시스템을 암호화하였는데 국내 ASD 로그에서도 동일한 유형의 공격 사례가 확인되었다.
Observables
141 indicators attributed across vendor reports, sandbox runs, and researcher write-ups. Values are masked here and available in full in Mallory.
Recent activity
7 sources tracked across advisories, community write-ups, and news. New activity surfaces here as Mallory finds it.