ThreatBeaconXSubscribe
HighAPT · 12 min read

DarkBlinders Espionage Campaign Uses GitHub-Based C2 to Target Government and Security Entities

A cyberespionage campaign attributed to the DarkBlinders threat actor has targeted government and security-sector entities in Israel and the Kurdistan Region of Iraq. Observed between August and October 2026, the campaign combines government-themed phishing pages, credential theft, deceptive cloud-drive links, and a fake video-conferencing application named StarkMeet to deliver a multi-stage Windows backdoor. The malware uses GitHub repositories as command-and-control infrastructure, separating initial host registration from second-stage activation. This approach allows operators to collect system information, assess infected devices, and selectively deploy a backdoor capable of executing PowerShell commands, transferring files, and collecting data. Researchers confirmed the compromise of a high-profile individual in Israel and a government cloud environment in the Kurdistan Region of Iraq. At least 1 GB of data was reportedly exfiltrated from the Iraqi government cloud environment. Infrastructure and malware correlations connect the activity to UNC5795 and Dust Specter, with a possible broader relationship to APT34. These attribution assessments carry varying confidence levels and should not be treated as definitive organizational identification.

Written by ThreatBeaconX Research Team·Published Nov 3, 2026

1. Executive Summary

Dream Research Labs identified an active cyberespionage operation conducted by DarkBlinders, the threat actor previously associated with the Blinder Tunnel campaign documented by Unit 42.

The campaign used multiple social-engineering approaches to compromise targeted users. These included counterfeit government webmail portals, cloud-drive sharing pages, and a deceptive video-meeting application called StarkMeet.

The investigation uncovered a multi-stage infection chain involving RuntimeBroker.dll and RuntimeBrokerApi.dll. The initial loader collects host metadata and registers infected systems with an attacker-controlled GitHub repository. The operator can then selectively provide activation material to decrypt and load the second-stage backdoor directly into memory.

The backdoor communicates through GitHub repositories and supports remote PowerShell execution, file uploads, file downloads, and archive extraction. A Cloudflare Worker provides an alternative communication route.

Key findings include:

  • Confirmed victim compromises: Researchers identified a compromised high-profile individual in Israel associated with the security sector and a government cloud environment in the Kurdistan Region of Iraq.

  • Data exfiltration: At least 1 GB of data was reportedly exfiltrated from the Iraqi government cloud environment.

  • Credential phishing: Fake Outlook Web App portals impersonated the Kuwait Ministry of Foreign Affairs and the Gulf Cooperation Council Secretariat General.

  • Malware delivery: StarkMeet was used as a deceptive meeting application to persuade victims to install malicious components.

  • Selective activation: The first-stage loader registered host information before the operator supplied the material needed to activate the second-stage backdoor.

  • GitHub-based C2: Repositories under the PeakyBlindersTeam account were used for host registration, command delivery, and result collection.

  • Campaign relationships: Infrastructure and malware overlaps connect the activity to UNC5795 and Dust Specter, with a possible relationship to UNC5187 and the broader APT34 cluster.

The campaign demonstrates how legitimate developer and cloud platforms can be repurposed for covert command-and-control operations, complicating detection based solely on network destinations.

2. Campaign Overview

AttributeDetails
Threat actorDarkBlinders
Related tracking namesUNC5795, Dust Specter
Possible broader associationAPT34
Activity observedAugust–October 2026
Primary targetsIsrael and Kurdistan Region of Iraq
Additional phishing themesKuwait and GCC government entities
Targeted sectorsGovernment, diplomacy, security
Primary platformWindows
Initial accessCredential phishing and deceptive application delivery
Malware componentsRuntimeBroker.dll, RuntimeBrokerApi.dll, StarkMeet
C2 infrastructureGitHub repositories and Cloudflare Worker
Confirmed impactTwo identified victims; at least 1 GB exfiltrated from an Iraqi government cloud environment

The operation combined credential theft with malware-based access. Some phishing pages were designed exclusively to capture credentials, while other delivery infrastructure presented malicious archives and applications.

The campaign's operational design separated host discovery from backdoor activation. This allowed the operator to review system information before deciding which infected devices should receive the second-stage payload.

3. Initial Access and Delivery Techniques

3.1 Government Webmail Credential Phishing

The campaign used counterfeit Outlook Web App login pages impersonating government entities.

Researchers recovered phishing pages imitating:

  • Kuwait Ministry of Foreign Affairs.

  • Gulf Cooperation Council Secretariat General.

The pages reproduced Microsoft Outlook Web App elements and referenced official-looking assets associated with government webmail services.

The credential forms submitted usernames and passwords to attacker-controlled endpoints, including paths under:

/owa/auth.owa

These pages were designed to collect credentials rather than exploit vulnerabilities or directly install malware.

Successful credential theft could provide access to email accounts, contacts, shared documents, and connected cloud services.

3.2 Cloud-Drive Delivery Lures

The operators also created fake cloud-drive pages resembling Google Drive sharing links.

The pages used familiar URL structures, including:

/drive/file/d/<identifier>/view

Examples included:

  • krgcloud[.]online

  • moelcloud[.]online

  • cdndrive[.]online

  • drive-g[.]cloud

One observed lure presented a ZIP archive named Attachment.zip. When the browser could not preview the archive, the page prompted the recipient to download it.

This technique uses a familiar file-sharing workflow to encourage users to download and execute attacker-provided content.

3.3 StarkMeet: Deceptive Video-Meeting Application

The campaign introduced a fake video-conferencing application named StarkMeet, presented as version 3.2.

The application displayed a conventional installation wizard and a functional-looking meeting interface. It included camera, microphone, and screen-sharing previews that operated locally.

However, the Join function consistently returned a connection failure rather than contacting a legitimate meeting service.

The visible application acted as a decoy while malicious components were installed separately under:

%LOCALAPPDATA%\Microsoft\RuntimeBroker

This separation allowed the malicious loader to remain persistent even if the visible StarkMeet application was removed.

The original message used to deliver StarkMeet.zip was not recovered. Therefore, the exact initial delivery mechanism for this archive remains unconfirmed.

4. Malware Execution and Infection Chain

The infection chain combines a deceptive application, a signed .NET host, an AppDomainManager-based loader, and an encrypted second-stage backdoor.

4.1 Initial Installation

The StarkMeet installer creates two distinct application locations:

LocationPurpose
%LOCALAPPDATA%\StarkMeetVisible meeting application
%LOCALAPPDATA%\Microsoft\RuntimeBrokerMalicious loader components

A legitimate signed executable, vshost.exe, is renamed to RuntimeBroker.exe and used to load RuntimeBroker.dll.

The loader checks for its expected execution path:

%LOCALAPPDATA%\Microsoft\RuntimeBroker\RuntimeBroker.exe

The use of a Microsoft-themed name and version information helps the malicious components blend into a Windows environment.

4.2 Persistence Through Registry Run Key

The malware establishes persistence using the following registry value:

HKCU\SOFTWARE\Microsoft\Windows\CurrentVersion\Run

Value name:

MicrosoftRuntime

The value points to the malicious RuntimeBroker.exe executable.

The first persistence check occurs approximately 121 seconds after startup and repeats hourly.

4.3 Host Identification and Registration

The loader generates a victim identifier using a SHA-256-based calculation involving:

  • User SID

  • Machine name

  • User domain

The identifier is derived from:

SHA256("Peaky Blinders 2.1" + UserSID + MachineName + UserDomainName)

The first 16 lowercase hexadecimal characters are used as the victim identifier.

The malware then uses an embedded GitHub personal access token to access the PeakyBlindersTeam/myLic repository.

Host information is uploaded under a victim-specific path:

{id}/{id}Inf.txt

The collected information includes:

  • Machine name and username.

  • Domain and logon server.

  • Time zone and local time.

  • Interface language and keyboard layouts.

  • Network interface information.

  • Persistence status.

  • Anti-analysis and virtual-machine findings.

The loader also checks for analysis tools, suspicious processes, virtual-machine artifacts, hardware characteristics, and other environmental indicators.

These findings are recorded and made available to the operator for host assessment.

4.4 Selective Second-Stage Activation

The initial loader does not automatically activate the full backdoor on every infected system.

Instead, it requests a host-specific activation file from the myLic repository:

{id}/{id}Lic.txt

The returned license string is used to decrypt RuntimeBrokerApi.dll.

The reported decryption process uses:

ParameterValue
AlgorithmAES-256-CBC
PaddingPKCS7
Key derivationSHA-256 hash of the license string
Initialization vectorFirst 16 bytes of the derived key
ExecutionDecrypted assembly loaded directly into memory

After obtaining the activation material, the loader decrypts the second-stage DLL and loads it into a new .NET AppDomain without writing the decrypted assembly to disk.

This design gives the operator control over which registered systems receive the fully functional backdoor.

4.5 GitHub-Based Command and Control

The second-stage component, RuntimeBrokerApi.dll, communicates with a separate GitHub repository:

PeakyBlindersTeam/myCode

The malware retrieves commands approximately every 63 seconds from a victim-specific command file.

The communication process involves:

  1. Retrieving tasking from the victim's repository directory.

  2. Reading and processing the issued command.

  3. Clearing the command file after retrieval.

  4. Executing the requested operation.

  5. Uploading the results to the same victim-specific repository directory.

Command and result data are Base64-encoded.

The malware also supports a Cloudflare Worker as a fallback communication relay. The Worker is validated using GitHub-like response characteristics before being accepted.

Important detection consideration: GitHub and Cloudflare are legitimate shared services. Organizations should not broadly block these platforms based only on this campaign. Detection should focus on the exact repository, account, Worker hostname, API paths, and associated process behavior.

4.6 Remote PowerShell Execution

The backdoor includes an embedded PowerShell execution helper named PsProxy.dll.

It loads the helper through .NET assembly loading and creates a PowerShell runspace using:

System.Management.Automation

Commands execute within the existing process rather than requiring a separate powershell.exe child process.

This can reduce visibility for detections that rely exclusively on monitoring PowerShell executable creation.

4.7 Supported Backdoor Commands

CommandFunctionDescription
/dlFile downloadRetrieves a file from the victim's repository directory and writes it locally
/upFile uploadUploads a local file to the victim's repository directory
/deArchive extractionDownloads and extracts a ZIP archive
/rateAPI statusQueries GitHub API rate-limit information

These functions provide the operator with remote file-management and data-transfer capabilities.

4.8 Dynamic C2 Configuration Through GitHub Comments

The malware includes a mechanism to replace its GitHub account, repository, and token configuration.

It periodically searches issues in the public Microsoft/vscode GitHub repository for comments containing a specially formatted hidden message.

The malware extracts the content from HTML comment markers, decodes and decrypts it, and checks for the required configuration fields:

  • Owner

  • LicRepo

  • LicToken

If all required fields are present, the malware updates its C2 configuration in memory.

This mechanism provides an alternative way to rotate communication infrastructure without directly modifying the malware on compromised systems.

5. Victimology and Confirmed Impact

Dream Research Labs identified approximately ten host registrations in the initial myLic repository. Only two systems were identified in the second-stage tasking channel.

The difference indicates that the operator may have selectively activated systems based on collected host information.

The two identified victims were:

  1. Government cloud environment in the Kurdistan Region of Iraq: The investigation identified post-compromise activity involving credential theft and the exfiltration of at least 1 GB of data.

  2. High-profile individual in Israel: The compromised individual was associated with the security sector.

The report also identified phishing infrastructure impersonating Kuwait's Ministry of Foreign Affairs and the GCC Secretariat General.

These phishing domains demonstrate broader targeting themes, but the presence of a lure does not independently establish a successful compromise of the impersonated organization.

6. Threat Attribution and Infrastructure Relationships

Dream assessed with medium-to-high confidence that the investigated DarkBlinders activity belongs to the same operational cluster as UNC5795 and Dust Specter.

The assessment is supported by:

  • Shared infrastructure and domain relationships.

  • Overlapping victimology.

  • Malware similarities across campaign waves.

  • Correlations involving HOTAIR, AEROSTAT, and TREEWORLD-related samples.

  • Reuse of infrastructure and operational patterns.

Researchers also identified an infrastructure connection between the UNC5795-associated activity and UNC5187. A server previously associated with UNC5187 malware was later linked to infrastructure used by the broader cluster.

The findings support a possible relationship between these activity clusters and APT34, but the precise organizational structure remains unresolved.

The report also identified contextual indicators of an Iranian nexus, including ParsVDS nameservers and a Persian keyboard layout observed on an early system registration. These observations are not sufficient on their own to establish the operator's location or identity.

Attribution should therefore remain qualified:

  • DarkBlinders to UNC5795/Dust Specter: medium-to-high confidence.

  • Relationship between UNC5795 and UNC5187: medium confidence.

  • Broader APT34 association: medium confidence.

7. Indicators of Compromise (IOCs)

The following indicators are extracted from the Dream Research Labs investigation. Domains, IP addresses, and URLs are defanged.

7.1 Domain Indicators

DomainObserved purpose
gcc-sg[.]cloudGCC Secretariat General webmail lure
gcc-sg[.]onlineGCC government-themed webmail lure
mfakuwait[.]orgKuwait Foreign Affairs phishing
mofakuwait[.]onlineKuwait Foreign Affairs lookalike
mfakuwait[.]onlineKuwait Foreign Affairs phishing
mfakuwait[.]websiteKuwait Foreign Affairs phishing
krgcloud[.]onlineKurdistan Regional Government cloud lure
moelcloud[.]onlineKRG Ministry of Electricity file lure
korektell[.]comKorek Telecom lookalike
msonedrive[.]cloudMicrosoft OneDrive lookalike
vision-cloud[.]onlineGeneric cloud-drive lure
gsecurity[.]helpGoogle account security lure
secure-googe[.]helpMisspelled Google security lure
googedrive[.]onlineGoogle Drive lookalike
cdn-gdrive[.]cloudGoogle Drive / CDN lure
cdndrive[.]onlineCloud-drive delivery lure
drivegooge[.]onlineGoogle Drive lookalike
drive-googe[.]onlineGoogle Drive lookalike
drive-g[.]cloudGoogle Drive-style file delivery
drive-go[.]camGoogle Drive-style lure
ukcentral[.]teamUnresolved UK-themed infrastructure
meetcloud[.]siteMeeting-service lure
meetgooge[.]onlineGoogle Meet lookalike
googemeet[.]onlineGoogle Meet lookalike
accountmeetgooge[.]onlineGoogle Meet account lure
meetgooge[.]camGoogle Meet lookalike
meet-goo[.]camGoogle Meet-style lure
meetlogin[.]onlineMeeting login lure
meetonline[.]camOnline meeting lure
stark-meet[.]comStarkMeet decoy and payload delivery
stark-meet[.]onlineStarkMeet-themed infrastructure
stark-meet[.]websiteStarkMeet-themed infrastructure
starkmeet[.]comStarkMeet landing and download site
artcyberspacetraining[.]siteUnresolved campaign infrastructure
optionbmitfilterport[.]siteUnresolved campaign infrastructure
artwavecollective[.]siteUnresolved campaign infrastructure
astravioncapital[.]siteUnresolved campaign infrastructure
lifeinmotionnow[.]siteUnresolved campaign infrastructure
g-prx.itugegape524.workers[.]devCloudflare Worker C2 proxy

7.2 IP Address Indicators

IP AddressPortObserved role
194.62.249[.]19—Google Drive and Meet lookalike hosting
194.62.249[.]1148001Google and webmail lookalikes; campaign service
194.62.249[.]53—Middle Eastern government and telecom lookalikes
89.125.209[.]80—StarkMeet-themed infrastructure
78.17.71[.]148—Vision-cloud infrastructure
77.223.215[.]140—Multi-service campaign host
51.79.96[.]115—Historical StarkMeet DNS resolution

7.3 URL Indicators

Credential phishing

hxxps://gcc-sg[.]cloud/owa/auth/logon.aspx?replaceCurrent=1&url=/owa/&reason=0

hxxps://mfakuwait[.]org/owa/auth/logon.aspx?replaceCurrent=1&url=/owa/&reason=0

hxxp://mfakuwait[.]online/owa/auth/logon.aspx?replaceCurrent=1

hxxps://mfakuwait[.]website/owa/auth/logon.aspx?replaceCurrent=1&url=/owa/&reason=0

File delivery

hxxps://cdndrive[.]online/drive/file/d/1JpH3ySmJpvX6TyGnF7C-S8qYiaoxyC5O/view

hxxps://drive-g[.]cloud/drive/file/d/15EMLDq9hpNXyZdu7RIZokPWLSdDC0y_0/view

hxxps://krgcloud[.]online/drive/file/d/1JpH3ySmJpvX6TyGnF7C-S8qYiaoxyC5O/view?ref=3yh3q5

hxxps://moelcloud[.]online/drive/file/d/1JpH3ySmJpvX6TyGnF7C-S8qYiaoxyC5O/view?ref=3yh3q5

Payload delivery

hxxps://stark-meet[.]com/wp-content/uploads/2025/08/StarkMeet.zip

hxxps://starkmeet[.]com/download

Operational infrastructure

hxxps://api.github[.]com/repos/PeakyBlindersTeam/myLic

hxxps://api.github[.]com/repos/PeakyBlindersTeam/myCode

hxxps://g-prx.itugegape524.workers[.]dev/rate_limit

hxxp://194.62.249[.]114/8001

Note: GitHub API endpoints and Cloudflare Worker infrastructure are hosted on shared services. Do not block github.com, api.github.com, workers.dev, or Cloudflare globally based only on these indicators. Prioritize exact repository, account, hostname, and URL-path matching.

7.4 SHA-256 File Hashes

SHA-256File / Description
052702d652286482863a70eaa37e4fc894f40c8b08dccf245ced46916e7f419fDubai Airport careers lure executable
06d68dcb8088546cb40fb3d7e87d6a593cbdf4cc1e43017604faf786ad6dbd41Malicious .NET payload
0702d54b03bf601019b6ad039f76ecc49b5b3e564119dfd60aebd742c8e90300Build artifact from lure package
0c9869b4aa7dde55936ab8726c442368b722eb69b785ef5c8e24938a01fdee89AppDomainManager loader configuration
11d1d570b5db7639d56a40621045e96c33b3d1ec3ae49e63c2ff07b9865844d4StarkMeet installer variant
13216e9942e15a48228005092f4e5ff8b7382d4ec73673f88c6d7b6c103cbce0Text file bundled with lure
3ab16d953c71738788ef2f39b62cf47284750e98aada273187954536e1ee1c83Shortcut launching hidden component
3f1961c951c105f289e64a8933c7d1ee34d71f1964999c01e97f2d944dfdc33cKuwait Foreign Affairs phishing document
3f76edab26d08b1517a15bde8ca82e94d48b4cbedd005eb11c950cfd6d9a8982Packed StarkMeet executable
3fd810a3aa0039993393741b32287c367a9a5037a41e826906440887cdd3ed13Embedded PowerShell helper
3fde5576233852cb9fdc0d6c2c4956a8eaa0857c2bbd86c388d7b4e10f11b0e8Kuwait Foreign Affairs credential page
415eb0d0055bc113df2da361b35c4de65ff55d5cddea3b637660b3ad19848ebdHTML and JavaScript phishing page
48e6a4196bff405b8b8ef80dcba599b93cb0428f3e587da90be6020e2337c83eStarkMeet landing page
53f35e49eb9b271fd8cbcd3daacb525328dbf159a03dbd1c7adebe0363daa402RuntimeBroker masquerading loader
6e7d9b33f1e72ea1ede71373a604ecdb060dab7d42055179c1eede9ecd1fd239Dubai Airport lure archive
770646093df203013b190ef0b0baa9dd2f6834534970a57ac5dc4be9b22ad28eKuwait Foreign Affairs Outlook credential page
824b36b9af74655a2c53b73959c144028b1f8ccc7c45ef17455d020b74cd0a51StarkMeet decoy executable
85c119297f42f09f9414e138b8f137a3e2dac1ee56fb6ae79c4c5044bc87bd80Campaign API JSON configuration
87bd8af14932f132bb3df9a6b8c17748c3e114432e217816a88ba35ec9a381f4StarkMeet executable variant
8875a41ead3b17eb8538466bfecb211a52d7ee56fa6e496b4407bdb0caa51dc5Signed host repurposed by loader
8c68119ebef2ec4be1d57f867ecc1e3606f819bd6278d35a95c6d9b789ac0bfdStarkMeet delivery archive
9ceb74f9ca2b0e9e1f0e579bfbb5d3d376667165eff9b8cda616b1c4a74236b9StarkMeet installer variant
aa7d4bf74edacba06da7e9d414c0649599fb0e7b118f7c3ddcec09eb3f720c82Alternate RuntimeBroker loader
c9178fef804f2d6a525c1af0a8ba49a5ec692bfaabd6043e5f0094fe93a31e82RuntimeBrokerApi.dll
aae2a7860f2e46019c38c1e7a50bc3b8f8104577acf750221c2cf23e6b376328StarkMeet installer variant
ac91ba73040ed42fb4eda1459bb8098e2f80970f2028da80c4e2f01e281cc223Shortcut launching hidden executable
b24a0c29134800dad72021e22d5ead99fa941722526807f29c131f6fbffa5fe2StarkMeet landing-site JavaScript
c039da6d430f78fc2a405c8ae945d2114518a63d73857145bb244ab651e41f84Second-stage StarkMeet executable
d8a4d54808c4705f472222c8a6ac5fd0622eb9bda7a2f07e12d712ea8b495d8eAlternate shortcut
dbd31b5d891d789e5cebbf0744a57b17c2f9a4fe50b79f64956e40dbb5fcaa9eOutlook Web App phishing page
e325ba0f93dfdeeb8c35e159c54f9d4e796905056506f6d627f4d66f15fb9988GCC Secretariat Outlook credential page
f2c54434e2eecca203c2114c0651e5d7a90fc2681117240686d084f016f9e2b6Alternate shortcut
f5b12772db6817f7a765a6fe7565fd3d4f87edc28e42fe3ec0244a372a410fc9Visual Studio project from lure archive

Additional hash: MD5 482c32947763a613199c43f7b4760f33 was listed in the source inventory without associated file context.

8. MITRE ATT&CK Mapping

TacticTechniqueID
Initial AccessSpearphishing LinkT1566.002
Initial AccessUser Execution: Malicious FileT1204.002
Credential AccessInput Capture: Web Portal CaptureT1056.003
ExecutionPowerShellT1059.001
PersistenceRegistry Run Keys / Startup FolderT1547.001
Defense EvasionObfuscated Files or InformationT1027
Defense EvasionVirtualization/Sandbox EvasionT1497
DiscoverySystem Information DiscoveryT1082
DiscoveryProcess DiscoveryT1057
DiscoverySystem Network Configuration DiscoveryT1016
Command and ControlWeb ServiceT1102
Command and ControlIngress Tool TransferT1105
CollectionData from Local SystemT1005
ExfiltrationExfiltration Over Web ServiceT1567
ExecutionShared ModulesT1129

These mappings represent observed or technically supported behaviors. Validate the exact technique subtypes against your environment and the full ATT&CK knowledge base before deploying detections.

9. Recommendations

Immediate Actions

  1. Search for known IOCs: Review DNS, proxy, firewall, EDR, email gateway, and SIEM logs for the listed domains, IP addresses, URLs, and SHA-256 hashes.

  2. Investigate suspicious RuntimeBroker execution: Review processes running from %LOCALAPPDATA%\Microsoft\RuntimeBroker and validate the digital signatures, parent processes, file hashes, and execution paths.

  3. Review registry persistence: Inspect HKCU\SOFTWARE\Microsoft\Windows\CurrentVersion\Run for the MicrosoftRuntime value and investigate any associated executable.

  4. Monitor GitHub-based C2 activity: Investigate unusual processes communicating with GitHub API endpoints, especially those accessing unfamiliar repositories or repeatedly exchanging encoded data.

  5. Review PowerShell runspace activity: Look for suspicious .NET applications loading System.Management.Automation and executing commands without creating a powershell.exe child process.

  6. Investigate phishing exposure: Search email and browser telemetry for StarkMeet-themed content, government webmail impersonation, and suspicious cloud-drive links.

  7. Validate credential compromise: If users submitted credentials to suspected phishing pages, reset affected passwords, revoke active sessions, and review authentication logs for unauthorized access.

  8. Assess data exposure: Investigate unusual cloud-file access, bulk downloads, archive creation, and outbound data transfers.

Preventive Actions

  1. Enforce phishing-resistant multifactor authentication for privileged and government-related accounts.

  2. Apply application control policies to restrict execution of untrusted installers and unsigned applications.

  3. Monitor unexpected .NET assembly loading and suspicious AppDomainManager-related execution patterns.

  4. Restrict unnecessary outbound access to developer platforms and cloud services through controlled proxy policies.

  5. Implement behavioral detections for unusual GitHub API activity originating from endpoint processes.

  6. Monitor registry Run-key modifications and validate newly registered startup applications.

  7. Train employees to verify government webmail and file-sharing domains before submitting credentials or downloading archives.

  8. Apply least-privilege controls and restrict access to sensitive cloud repositories and shared drives.

  9. Enable detailed endpoint, PowerShell, identity, and cloud audit logging.

  10. Maintain network segmentation and data-loss monitoring for sensitive government and security environments.

10. Conclusion

The DarkBlinders campaign demonstrates a structured espionage operation that combines credential phishing, deceptive application delivery, selective malware activation, and legitimate cloud-platform abuse.

Its GitHub-based command-and-control design enables the operator to register infected systems, review host information, and activate a second-stage backdoor only on selected targets. The backdoor's in-memory PowerShell execution and file-transfer capabilities provide a flexible mechanism for remote operations and data collection.

The confirmed compromise of a government cloud environment in the Kurdistan Region of Iraq, including the reported exfiltration of at least 1 GB of data, highlights the operational impact of this campaign.

Infrastructure and malware correlations indicate relationships with UNC5795 and Dust Specter, while a broader APT34 association remains an assessment rather than a definitive attribution.

Organizations should prioritize detection of suspicious RuntimeBroker execution, GitHub-based C2 behavior, unauthorized PowerShell runspaces, credential phishing, and unusual cloud-data access.

MITRE ATT&CK Mapping

T1566.002T1204.002T1056.003T1059.001T1547.001T1027T1082T1057T1016T1102T1105T1005T1567