← All research

Transparent Tribe APT36 Threat Actor Hunting Research

Analysis of Transparent Tribe threat actor profile, attack flows, indicators, campaigns, tooling, and hunting hypotheses for Microsoft Defender XDR and Sentinel

Executive Summary

Transparent Tribe is a persistent cyber-espionage actor active since at least 2013. MITRE ATT&CK tracks the group as Transparent Tribe (G0134) and associates the names APT36, COPPER FIELDSTONE, Mythic Leopard, and ProjectM with it. Public reporting assesses the actor as Pakistan-based or Pakistan-aligned with high confidence, but does not conclusively identify a sponsoring government agency.

The actor primarily targets Indian military, defense, government, diplomatic, aerospace, research, and education organizations. Afghanistan, regional officials, activists, and diplomatic targets also appear in public reporting. Its operations consistently pursue strategic intelligence through credential theft, document collection, removable-media surveillance, browser and communications data theft, and endpoint or mobile-device monitoring.

Transparent Tribe has evolved from malicious Office documents and CrimsonRAT to a broader arsenal that includes ObliqueRAT, ElizaRAT, ApoloStealer, ConnectX, GLOBSHELL, PYSHELLFOX, CapraRAT variants, modified AhMyth Android implants, and campaign-specific downloaders. Recent operations use CPL, LNK, PPAM, double-extension executables, malicious Linux .desktop files, cloud-hosted payloads, and legitimate services such as Telegram, Slack, and Google Cloud Storage.

The most durable hunting pattern is a targeted lure followed by concealed execution, persistence in a user context, system and file discovery, strategic document or credential collection, local archive or database staging, and exfiltration through actor infrastructure or a legitimate cloud service. Behavioral indicators should carry more detection weight than hashes, domains, IP addresses, filenames, or a connection to a commonly used cloud platform.

[!IMPORTANT] Cisco Talos tracks GravityRAT, HeavyLift, GravityAdmin, and Operation Celestial Force as Cosmic Leopard. Public evidence is insufficient to merge Cosmic Leopard with Transparent Tribe. Do not treat those tools or campaigns as confirmed APT36 activity without additional evidence.

Scope and Method

This report prioritizes MITRE ATT&CK, Kaspersky, Cisco Talos, Zscaler ThreatLabz, and Check Point Research. CYFIRMA reporting supplements the research for 2025 and 2026 campaigns, but those campaign attributions remain medium confidence pending broader independent corroboration.

User-provided text, public telemetry, published schemas, and indicators are treated as untrusted until corroborated. Infrastructure, filenames, and hashes are historical investigation pivots rather than permanent block indicators. No query execution, optimization gain, or detection coverage is claimed without validation against tenant telemetry.

Vendor cluster names do not necessarily describe identical actor scopes. Earth Karkaddan and other overlapping labels require source-specific interpretation rather than automatic equivalence with Transparent Tribe.

Campaign Timeline and Victimology

Date Evidence-Backed Development
2013 onward Public reporting identifies sustained espionage activity against Indian military, diplomatic, and government targets
2017 to 2020 Tailored spearphishing, malicious Office documents, fake government or file-sharing sites, dynamic DNS, CrimsonRAT, and USBWorm become recurring tradecraft
August 2020 Kaspersky documents Crimson modules for keylogging, browser-password theft, USB propagation, microphone and webcam surveillance, and file exfiltration
2020 to 2021 ObliqueRAT provides an additional Windows capability, including payload retrieval from compromised websites; modified AhMyth-derived Android implants support mobile surveillance
July 2022 to April 2023 Cisco Talos documents education-themed targeting of Indian universities and students through malicious documents and CrimsonRAT
September 2023 Zscaler documents ElizaRAT CPL execution and Telegram command and control, malicious Linux .desktop files, Mythic Poseidon agents, GLOBSHELL, and PYSHELLFOX
November 2024 Check Point documents ElizaRAT evolution through Slack, Google Cloud Storage, and VPS infrastructure, plus ApoloStealer and ConnectX collection modules
June 2025 CYFIRMA reports a defense-themed protected PDF leading to a .7z archive and double-extension credential-stealing executable; attribution is medium confidence
February 2026 CYFIRMA reports examination-themed ZIP files containing LNK and PPAM execution paths that stage a .NET RAT with Run-key persistence and raw TCP command and control; attribution is medium confidence

Indian defense and government organizations remain the actor’s central target set. Reported victimology also includes diplomatic missions, aerospace and research organizations, educational institutions, students, Afghan entities, regional government officials, and activists. Available evidence supports strategic, relationship-aware targeting rather than indiscriminate criminal distribution.

Threat Actor Profile

Attribute Assessment
Motivation Strategic cyber espionage with high confidence; financial or destructive objectives are not prominent
Assessed origin Pakistan-based or Pakistan-aligned with high confidence; Zscaler states very high confidence
Sponsorship caveat Public evidence does not conclusively identify a sponsoring Pakistani government or intelligence agency
Primary targeting Indian military, defense, government, diplomatic, aerospace, research, and education organizations
Additional targeting Afghanistan, regional officials, activists, government-adjacent organizations, and occasional diplomatic targets elsewhere
Collection priorities Credentials, government and military documents, email and browser sessions, removable-media files, communications, screenshots, audio, video, location, and mobile data
Access model Tailored spearphishing, malicious links and attachments, cloud-hosted archives, fake portals, social engineering, and trojanized mobile applications
Infrastructure model Dynamic DNS, compromised websites, virtual private servers, actor domains, and legitimate cloud or collaboration services
Attribution confidence High confidence in the long-running Transparent Tribe activity cluster; medium confidence in several 2025 and 2026 campaign-specific links

Attribution should rest on the full chain of victimology, lure context, execution, persistence, collection behavior, infrastructure, and malware lineage. A single RAT family, cloud service, filename, or network indicator does not prove Transparent Tribe involvement.

Infection and Attack Flow

flowchart TD
    A[Targeted Email, Message, or Relationship Lure] --> B{Delivery Format}
    B --> C[Office Document, Archive, LNK, CPL, or PPAM]
    B --> D[Linux Desktop File Masquerading as a Document]
    B --> E[Trojanized Android Application]
    C --> F[Macro, Script, Rundll32, or User Execution]
    D --> G[Shell, Wget, Chmod, and ELF Execution]
    E --> H[Permission Requests and Secondary APK]
    F --> I[CrimsonRAT, ObliqueRAT, ElizaRAT, or .NET RAT]
    G --> J[Poseidon or Python Collection Utility]
    H --> K[CapraRAT or AhMyth-Derived Implant]
    I --> L[Run Key, Startup LNK, or Scheduled Task]
    J --> M[User Crontab and Hidden Directory]
    K --> N[Mobile Surveillance and File Collection]
    L --> O[Discovery, Credential Access, and Collection]
    M --> O
    N --> P[HTTP Command and Control]
    O --> Q[ZIP, SQLite, Hidden Path, or Memory-Backed Staging]
    Q --> R[Actor Server, Telegram, Slack, Google, or VPS Exfiltration]
    P --> R

Stage 1: Targeting and Delivery

Transparent Tribe tailors lures to military, government, diplomatic, education, conference, résumé, examination, and relationship themes. Delivery formats include malicious Office documents, password-protected archives, PDF links, LNK files, CPL files, PPAM add-ins, double-extension executables, Linux .desktop files, fake government portals, and trojanized Android applications.

Cloud storage and compromised websites frequently host payloads or archives. The use of a legitimate hosting service is an evasion and delivery choice, not an attribution signal by itself.

Stage 2: Execution and Decoy Activity

Windows chains use Office macros, OLE content, batch files, VBA, LNK files, Control Panel items, rundll32.exe, or direct user execution. ElizaRAT campaigns commonly combine CPL execution with a decoy document or video and create a working directory under AppData. The reported 2026 chain uses LNK and PPAM paths, batch and VBA staging, Mark-of-the-Web removal, and a hard link before .NET RAT execution.

Linux campaigns place a padded .desktop file in an archive and give it a document icon. Its Exec= field invokes a shell, downloads a decoy and ELF payload with wget, writes into /tmp or ~/.local/share, changes permissions, and executes the implant.

Stage 3: Persistence and Command and Control

Windows persistence includes Registry Run keys, Startup-folder LNK files, and scheduled tasks. Linux campaigns use user crontabs and hidden paths. Android persistence and collection depend on the permissions and services implemented by each trojanized application.

Command-and-control channels include HTTP, HTTPS, raw TCP, Telegram, Slack, Google services, dynamic DNS, compromised sites, and actor-controlled VPS infrastructure. Cloud APIs can carry victim registration, commands, payload downloads, and exfiltrated files.

Stage 4: Discovery and Credential Access

Observed malware performs host, user, process, filesystem, and environment discovery. Crimson modules and later campaign payloads support keylogging and browser credential access. Linux tooling has targeted Firefox sessions and Indian government webmail. Reported campaigns also inspect machine names, time zones, and execution environments to identify analysis systems or unsuitable targets.

Stage 5: Collection and Exfiltration

Collection focuses on office documents, PDFs, archives, browser sessions, email-related data, WhatsApp-related data, screenshots, clipboard content, removable-media files, audio, video, and mobile communications. ApoloStealer performs broad document collection, while ConnectX focuses on USB media. SQLite databases, ZIP archives, hidden directories, and /dev/shm support local staging before transfer through the active command-and-control channel or cloud service.

Commonly Observed Tools and Capabilities

Tool or Capability Role and Observed Use
CrimsonRAT Windows remote access, command execution, discovery, file transfer, and modular surveillance
USBWorm Crimson-associated removable-media propagation and directory-lookalike behavior
ObliqueRAT Windows espionage implant with payload delivery through attacker or compromised infrastructure
ElizaRAT Windows implant using CPL-based delivery and Telegram, Slack, Google, or VPS command and control
ApoloStealer Broad document discovery, collection, SQLite staging, and exfiltration
ConnectX Removable-media monitoring and USB file collection
GLOBSHELL Windows collection and exfiltration capability documented in the 2023 arsenal
PYSHELLFOX Python-based collection focused on documents, browser sessions, webmail, and communications data
Mythic Poseidon Third-party Mythic agent observed in Linux operations; its presence alone does not identify APT36
CapraRAT variants Android surveillance implants associated with Transparent Tribe mobile operations
Modified AhMyth implants Android implants collecting SMS, contacts, calls, location, audio, WhatsApp media, and files
Peppy Malware associated with Transparent Tribe in public reporting
Living-off-the-land utilities rundll32, shell interpreters, wget, chmod, archive tools, Registry utilities, scheduled tasks, and crontab
Hosting and C2 services Dynamic DNS, compromised websites, Telegram, Slack, Google Drive or Cloud Storage, and VPS infrastructure

MITRE ATT&CK Map for Hunting

Technique ID Confidence Evidence Pattern to Hunt
Spearphishing Attachment T1566.001 High Targeted Office document, archive, LNK, CPL, PPAM, or executable attachment
Spearphishing Link T1566.002 High Targeted link leading to a fake portal, protected archive, or hosted payload
Malicious File T1204.002 High User opens an archive, document, control panel item, desktop entry, or APK
Windows Command Shell T1059.003 High Batch or command-shell staging and payload execution
Unix Shell T1059.004 High .desktop entry invokes bash or sh for download and execution
Visual Basic T1059.005 High Office VBA and PPAM execution paths stage payloads
Control Panel T1218.002 High rundll32 or Control Panel executes a CPL from a user-writable path
Registry Run Keys and Startup Folder T1547.001 High Run-key values or Startup LNK files launch the implant
Scheduled Task or Job T1053 High Windows scheduled task or Linux user crontab provides persistence
Masquerading T1036 High Double extensions, document icons, fake portals, or service-like filenames conceal payloads
Obfuscated Files or Information T1027 High Protected archives, obfuscated .NET or VBA, and heavily padded .desktop files
System Information Discovery T1082 High Implant collects host, environment, time-zone, or machine-name data
Process Discovery T1057 High Implant inventories running processes before collection or tasking
File and Directory Discovery T1083 High Recursive discovery identifies strategic files and removable-media content
Credentials from Web Browsers T1555.003 High Crimson and later payloads access browser credentials or session data
Input Capture: Keylogging T1056.001 High Crimson modules and reported later payloads record keyboard input
Screen Capture T1113 High Windows and mobile implants collect screenshots
Audio Capture T1123 High Crimson and Android variants collect microphone audio
Video Capture T1125 High Crimson surveillance modules access webcam video
Data from Removable Media T1025 High Crimson, ConnectX, and collection utilities target USB files
Local Data Staging T1074.001 High SQLite, ZIP, hidden directories, or /dev/shm stage collected data
Web Protocols T1071.001 High HTTP, HTTPS, and cloud APIs carry commands, payloads, or files
Non-Application Layer Protocol T1095 High Crimson and the reported 2026 RAT use raw TCP command and control
Exfiltration to Web Service T1567 High Telegram, Slack, or Google services transfer commands or collected files
Exfiltration Over C2 Channel T1041 High Implants upload documents, archives, databases, and surveillance output
Replication Through Removable Media T1091 High USBWorm copies itself and substitutes lookalike entries on removable media
Virtualization or Sandbox Evasion T1497 Medium-high Machine-name, time-zone, delay, and environment checks reduce analysis exposure

Mappings that depend only on a vendor technique list, without demonstrated behavior, are excluded. Public evidence does not support adding bootkit, rootkit, pre-OS, privilege-escalation, or exploitation techniques to the general APT36 profile without campaign-specific proof.

Durable Indicators of Attack

Attack Phase Durable Behavioral Indicator Hunting Value
Delivery Targeted defense, government, education, résumé, conference, examination, or relationship lure contains a protected archive or unusual executable format Narrows investigation when combined with recipient role and attachment ancestry
Execution Office, shell, LNK, PPAM, or archive activity creates and runs content from AppData, ProgramData, Downloads, /tmp, or ~/.local/share Stronger than a campaign filename because paths and parent-child relationships persist
Signed binary proxy execution rundll32.exe or Control Panel loads a CPL or DLL from a user-writable path High-value ElizaRAT-compatible behavior with manageable environmental tuning
Linux execution Document-themed .desktop file invokes a shell, wget, chmod, and an ELF payload Unusual and comparatively discriminating outside software installation workflows
Persistence Run key, Startup LNK, scheduled task, or user crontab references a recent file in a user-writable or hidden location Connects initial execution to durable access
Collection Unsigned or rare process recursively enumerates strategic documents, browser profiles, or removable drives Captures multiple malware families through shared collection objectives
Staging Document or USB enumeration is followed by ZIP, SQLite, hidden-directory, or /dev/shm staging Connects discovery to likely exfiltration preparation
Cloud C2 Uncommon process makes periodic API calls to Telegram, Slack, or Google and then writes or uploads files Detects abuse of trusted services without blocking the service globally
Surveillance Repeated screenshots, keylogging, clipboard access, or browser-session access follows initial execution Supports espionage-focused triage and scoping
Exfiltration Staged archives or databases are followed by repeated small uploads or transfer to a rare VPS Connects collection to confirmed data movement

Volatile Indicators of Compromise

Indicators in this section are historical enrichment. Confirm ownership, reputation, observation time, and local context before blocking. Shared hosting, cloud services, and reassigned infrastructure can create unrelated matches.

Network Indicators

Type Indicator Campaign Context Source Date
IP address 64.188.25.206 CrimsonRAT command and control August 2020
IP address 173.212.192.229 CrimsonRAT command and control August 2020
IP address 45.77.246.69 CrimsonRAT command and control August 2020
Domain newsbizupdates.net CrimsonRAT command and control August 2020
Domain uronlinestores.net CrimsonRAT command and control August 2020
Domain tryanotherhorse.com Android configuration and command and control August 2020
IP address 212.8.240.221 Android and ObliqueRAT-related infrastructure August 2020
Domain studentsportal.live Education-themed lure and infrastructure July 2022 to April 2023
IP address 198.37.123.126 Education-themed campaign infrastructure July 2022 to April 2023
Domain richa-sharma.ddns.net Education-themed CrimsonRAT infrastructure July 2022 to April 2023
Domain email9ov.in ElizaRAT campaign infrastructure September 2023
Domain baseuploads.com ElizaRAT and Linux payload hosting September 2023
IP address 103.2.232.82 ElizaRAT and Linux campaign infrastructure September 2023
IP address 38.54.84.83 Circle campaign command and control November 2024
IP address 83.171.248.67 Slack-themed ElizaRAT campaign command and control November 2024
Domain superprimeservices.com Defense-themed PDF-to-archive delivery June 2025
IP address 93.127.130.89 Reported raw TCP command and control February 2026
Domain sharemxme126.net Reported fallback command-and-control domain February 2026

File Indicators

Hash Type Hash Context Source Date
MD5 0294f46d0e8cb5377f97b49ea3593c25 Android dropper August 2020
MD5 d7d6889bfa96724f7b3f951bc06e8c02 ObliqueRAT-related dropper August 2020
SHA-1 516db7998e3bf46858352697c1f103ef456f2e8e Education-themed CrimsonRAT sample July 2022 to April 2023
MD5 fc99daa2e1b47bae4be51e5e59aef1f0 ElizaRAT CPL September 2023
MD5 65167974b397493fce320005916a13e9 Malicious Linux .desktop file September 2023
SHA-256 06d9662572a47d31a51adf1e0085278e0233e4299e0d7477e5e4a3a328dea9d1 ElizaRAT dropper November 2024
SHA-256 d66ba4ee97a2f42d85ca383f3f61a2fac4f0b374aad1337f5f29245242f2d990 ApoloStealer November 2024
SHA-256 f03ac870cb91c00b51ddf29b6028d9ddf42477970eafa7c556e3a3d74ada25c9 Reported defense-phishing payload June 2025
SHA-256 34412e765822cf3fb32a5a5c9866fb29a9b98d627b4d9a3275fd3e754cf8e360 Reported multi-vector Windows campaign component February 2026

Observable Signals by Attack Phase

Phase Observable Signals Defender XDR and Sentinel Sources
Delivery Targeted message, cloud-hosted archive, protected attachment, suspicious link, or uncommon file type EmailEvents, EmailAttachmentInfo, EmailUrlInfo, UrlClickEvents, mail gateway, proxy, and DNS logs
Windows execution Office, archive utility, LNK, PPAM, or rundll32.exe creates or runs content from a user-writable path DeviceProcessEvents, DeviceFileEvents, DeviceImageLoadEvents, DeviceEvents
Linux execution .desktop file launches shell, wget, chmod, or an ELF payload from a temporary or hidden path DeviceProcessEvents, DeviceFileEvents, Linux audit logs, proxy, and DNS logs
Persistence New Run key, Startup LNK, scheduled task, or user crontab references a recent payload DeviceRegistryEvents, DeviceFileEvents, DeviceProcessEvents, DeviceEvents, Linux audit logs
Discovery Rare process inventories systems, processes, files, browser profiles, or removable media DeviceProcessEvents, DeviceFileEvents, DeviceEvents
Credential access Browser data access, keylogging behavior, suspicious profile database reads, or credential-themed fake portal activity DeviceFileEvents, DeviceProcessEvents, browser, proxy, identity, and sign-in logs
Collection Document extension filtering, recursive USB access, screenshots, or surveillance data generation DeviceFileEvents, DeviceEvents, DeviceProcessEvents, mobile threat defense telemetry
Staging ZIP or SQLite creation, hidden-directory staging, or writes to /dev/shm after collection DeviceFileEvents, DeviceProcessEvents, Linux audit logs
Command and control Rare process contacts actor VPS infrastructure or legitimate cloud APIs at a periodic cadence DeviceNetworkEvents, proxy, DNS, firewall, and CloudAppEvents
Exfiltration Archive or database staging is followed by repeated uploads or unusual cloud egress DeviceNetworkEvents, proxy, firewall, and cloud application logs

Standard Defender XDR endpoint coverage does not provide complete Android surveillance visibility. Mobile hunting requires mobile threat defense, application, network, or device-management telemetry appropriate to the deployed platform.

Priority Hunting Hypotheses

ID Hypothesis Correlation Logic Candidate Predicates and Telemetry Priority False Positives and Data Gaps
H1 A targeted Office or archive lure deploys a Transparent Tribe-compatible Windows implant External lure is followed by Office or archive execution, payload creation in AppData or ProgramData, persistence, and outbound command and control Email and URL telemetry joined to DeviceProcessEvents, DeviceFileEvents, DeviceRegistryEvents, and DeviceNetworkEvents Critical Trusted macros, packaged applications, and software deployment can resemble individual steps; recover message and file ancestry
H2 A CPL or DLL from a user-writable path launches ElizaRAT-compatible activity Downloaded archive produces a CPL or DLL, rundll32.exe loads it, a decoy appears, and a Startup LNK or task precedes cloud or VPS traffic rundll32.exe, .cpl or .dll path, image load, recent download, Startup file, scheduled task, cloud API, and rare destination Critical Legitimate control-panel software requires signer, path, prevalence, and installation-context baselines
H3 A malicious Linux desktop entry installs a persistent espionage implant Archive extraction is followed by .desktop execution, shell, wget, chmod, ELF launch, hidden file writes, and user crontab modification Linux process and file telemetry, .desktop Exec= content, /tmp, ~/.local/share, /dev/shm, crontab, DNS, and proxy logs Critical User-installed desktop applications can use similar commands; document icon, download ancestry, and hidden staging increase confidence
H4 A rare process stages strategic documents or removable-media files for exfiltration USB insertion or recursive document enumeration is followed by ZIP or SQLite creation and outbound transfer DeviceEvents, DeviceFileEvents, DeviceProcessEvents, removable-media identifiers, archive/database creation, and DeviceNetworkEvents High Backup, indexing, DLP, and synchronization tools require process and account allowlists
H5 A nonstandard process uses a collaboration or cloud API as command and control Rare unsigned process establishes periodic cloud API traffic, writes a downloaded file, executes it, or uploads staged content Process signer and prevalence, API hostname and path, connection cadence, file writes, child execution, DeviceNetworkEvents, proxy, and CloudAppEvents High Approved bots, browsers, collaboration clients, and automation require application and account baselines
H6 A credential-themed government or defense lure leads to account compromise and endpoint persistence Lookalike portal or protected document precedes credential submission, anomalous sign-in, browser data access, keylogging, or RAT persistence Email and click telemetry, proxy redirects, domain age, identity sign-ins, IdentityLogonEvents, endpoint file and process activity Critical Legitimate government portals and password-reset workflows require domain verification and redirect-chain analysis
H7 A user-context persistence artifact launches a recently downloaded rare binary Run key, Startup LNK, task, or crontab references a file created shortly after email, browser, archive, or shell activity DeviceRegistryEvents, DeviceFileEvents, DeviceProcessEvents, task telemetry, Linux audit logs, file prevalence, and signer High Many legitimate applications persist in user context; recent-download ancestry and rare network traffic reduce noise
H8 Surveillance and browser-session collection follow a targeted lure Initial execution is followed by recurring screenshots, browser-profile reads, clipboard or input capture, and periodic small uploads Screenshot-file patterns, browser database access, API behavior, process modules, DeviceFileEvents, DeviceProcessEvents, and DeviceNetworkEvents High Accessibility, support, password-management, and browser-security products require process baselines and sequence correlation

Query-Building Blocks

Building Block Detection Intent Candidate Predicates
Targeted delivery Identify role-aware lures and uncommon executable content External sender, protected archive, suspicious link, defense or education theme, LNK, CPL, PPAM, EXE, APK, or .desktop
User-writable proxy execution Detect ElizaRAT-compatible CPL or DLL launch rundll32.exe, Control Panel item, Downloads or AppData path, rare hash, unsigned image, recent archive extraction
Linux masquerading Detect a document-themed desktop entry that invokes a shell .desktop, document icon, Exec=, bash, sh, wget, /tmp, ~/.local/share, chmod, or ELF child
User-context persistence Detect persistence tied to a recent payload Run key, Startup LNK, scheduled task, crontab, hidden path, recent file creation, rare signer or hash
Strategic collection Detect actor objectives rather than a specific malware family Recursive office-document access, browser profile reads, USB enumeration, screenshots, clipboard or keylogging artifacts
Local staging Detect preparation for espionage exfiltration ZIP, SQLite, hidden directory, /dev/shm, extension-filtered collection, archive or database growth
Cloud-service command and control Detect unusual API usage without globally blocking a service Rare process, direct API connection, periodic cadence, file download, child execution, upload, secondary VPS contact
Credential lure correlation Connect phishing, endpoint, and identity compromise Lookalike domain, redirect chain, credential submission, unusual sign-in, new device, token activity, endpoint persistence

Useful correlation entities include DeviceId, normalized hostname, recipient address, sender domain, AccountSid, InitiatingProcessAccountSid, source and destination IP addresses, URL, file hash, removable-media identifier, and cloud application identity. Candidate time windows range from one hour for execution chains to 24 hours for delivery-to-persistence and identity correlation. These windows require tenant-specific validation.

Detection Engineering Guidance

  • Prioritize behavior across delivery, execution, persistence, collection, staging, and exfiltration over static indicators
  • Correlate recipient role and lure theme with endpoint ancestry instead of relying on keywords alone
  • Baseline legitimate rundll32.exe control-panel activity by path, signer, hash prevalence, and installer ancestry
  • Inspect the Exec= value of suspicious Linux .desktop files and retain extracted archives for analysis
  • Detect direct cloud API access from rare binaries while excluding approved browsers, collaboration clients, and automation
  • Correlate removable-media access with archive or SQLite staging and subsequent egress
  • Treat a cloud-service hostname, third-party Mythic agent, or dual-use utility as supporting evidence rather than attribution proof
  • Use historical domains, IP addresses, and hashes for enrichment, scoping, and retrospective search rather than standalone high-severity alerts
  • Validate table and column availability against the target tenant before implementing queries or analytic rules
  • Tune Android hypotheses to the mobile threat defense and device-management data actually available

An example evidence score for prototyping is: targeted lure +2, suspicious execution +3, user-context persistence +3, strategic collection +3, local staging +2, unusual cloud or VPS command and control +3, and confirmed exfiltration +4. This score is a design aid and requires tenant-specific testing before production use.

Triage and Containment Guidance

Phase Recommended Actions
Confirm Recover the original message, URL chain, archive, decoy, execution ancestry, persistence artifact, collection output, and network timeline
Preserve Capture memory where appropriate, original payloads, LNK or CPL content, PPAM and VBA, .desktop Exec= values, crontabs, Run keys, tasks, cloud API requests, and proxy or DNS logs
Scope Search for the same sender, lure, hash, signer, filename, persistence value, destination, cloud API pattern, removable-media behavior, and staged archive across the environment
Contain Isolate affected endpoints, remove malicious persistence after evidence capture, block confirmed actor infrastructure, revoke exposed cloud tokens, and restrict unauthorized cloud API access
Credential response Reset exposed credentials, revoke sessions and tokens, inspect browser-stored credentials, and investigate sign-ins from new locations, devices, or infrastructure
Mobile response Remove trojanized applications, revoke application permissions, inspect secondary APK installation, rotate credentials used on the device, and follow mobile threat defense guidance
Recovery Reimage systems when implant removal cannot be verified, restore trusted application settings, and monitor for recurring persistence or command-and-control activity

Research Gaps and Confidence Limits

  • Public reporting does not conclusively identify a sponsoring Pakistani government or intelligence agency
  • Vendor actor clusters and aliases are not proven to be perfectly coextensive
  • The 2025 and 2026 campaign attributions have limited independent corroboration and remain medium confidence
  • Public victim counts and complete sector distributions are unavailable
  • Delivery and post-compromise details are incomplete for several malware samples
  • Android behavior varies materially by application and implant version, and standard endpoint telemetry may not observe it
  • Legitimate cloud services, compromised sites, dynamic DNS, and VPS infrastructure can be shared with unrelated activity
  • IoCs may expire, be reassigned, or appear in unrelated investigations
  • Public reports do not establish that every listed capability appears in every campaign or intrusion

Source References

Implementation Next Step

Convert hypotheses H1 through H8 into versioned Microsoft Defender XDR and Sentinel hunting queries. Validate table and column availability against the target tenant, then test expected results, false-positive rates, entity normalization, and cross-table time windows before promoting any query to a detection rule.