Back to Intelligence

Mistic Backdoor, Autonomous AI Agents, and GitHub Actions Abuse: OTX Pulse Analysis — Enterprise Detection Pack

SA
Security Arsenal Team
July 24, 2026
6 min read

Threat Summary

The latest OTX pulses reveal a convergence of advanced delivery mechanisms and traditional credential theft operations. A notable threat actor, Woodgnat, is actively deploying a stealthy backdoor named Mistic, often alongside ModeloRAT, to facilitate initial access for various ransomware operations (Qilin, Akira, Black Basta). Simultaneously, a sophisticated espionage campaign targeting Thailand's Ministry of Finance utilizes an autonomous Hermes AI agent operating in "YOLO" mode to stage the Hades implant. On the supply chain front, threat actors are weaponizing the AI ecosystem via ClawHub and compromising GitHub Actions to distribute malware and exploit cPanel/WHM vulnerabilities (CVE-2026-41940). Collectively, these campaigns aim to steal credentials, establish persistent backdoors, and leverage compromised infrastructure for lateral movement.

Threat Actor / Malware Profile

Woodgnat / Mistic Backdoor

  • Distribution: Social engineering, sideloading DLLs via legitimate-looking applications (e.g., fake login portals).
  • Payload Behavior: Establishes C2 channels via specific domains (e.g., authorized-logins.net). Often used in conjunction with loaders like MintsLoader and D3F@ck Loader.
  • Persistence: Likely utilizes DLL side-loading to maintain persistence under the guise of legitimate system processes.
  • Objective: Initial access brokering for ransomware affiliates.

Hades Implant & Hermes AI Agent

  • Distribution: Exploitation of vulnerabilities in Apache Hadoop (HiveServer2) and other web services (CVE-2026-43500, CVE-2026-43284).
  • Payload Behavior: A custom Go-based implant (Hades) deployed by an autonomous AI agent (Hermes). Capable of downloading further modules and stealing credentials.
  • C2 Communication: Communicates with attacker-controlled infrastructure to receive commands and exfiltrate data.

ClawHavoc & AI Supply Chain

  • Distribution: Malicious "skills" uploaded to the ClawHub marketplace (OpenClaw), specifically targeting AI agent workflows.
  • Techniques: File padding to bypass scanner thresholds and evasion of VirusTotal/ClawScan detection.
  • Objective: Infostealing (AMOS) and financial gain via agentic threats.

IOC Analysis

The provided indicators span multiple infrastructure types:

  • Domains & Hostnames: Numerous C2 domains such as mail.authorized-logins.net, mueleer.com, and laosji.net. SOC teams should immediately block these at the perimeter and DNS layer.
  • File Hashes (SHA256): Payloads for Mistic, Hades, and the AI supply chain malware (e.g., 3f797a639bc855bc6d5471f327924b62d10900ddec49b970eca6604142bbb4be). These should be added to EDR blocklists.
  • IPv4 Addresses: Scanner and C2 IPs like 43.246.208.207 and 43.228.157.68. These are critical for firewall blacklisting.
  • CVEs: Exploits for CVE-2026-41940 (cPanel), CVE-2017-7269 (IIS 6.0), and Linux privilege escalation (CVE-2021-4034).
  • URLs: Endpoint URLs used in the GitHub Actions campaign (http://43.228.157.68:80/api/dl/amd64).

Detection Engineering

Sigma Rules

YAML
title: Potential Mistic Backdoor or ModeloRAT Sideloading Activity
id: 8a7b8c9d-0e1f-2a3b-4c5d-6e7f8a9b0c1d
description: Detects potential sideloading behavior associated with Mistic backdoor or ModeloRAT based on suspicious DLL loads by legitimate Windows binaries.
status: experimental
date: 2026/07/24
author: Security Arsenal
references:
    - https://otx.alienvault.com/pulse/66a1c5b8e7b2d
logsource:
    category: image_load
    product: windows
detection:
    selection:
        ImageLoaded|contains:
            - '\AppData\Local\Temp\'
            - '\ProgramData\'
        Image|endswith:
            - '\rundll32.exe'
            - '\regsvr32.exe'
            - '\svchost.exe'
    filter_legitimate:
        Signed: 'true'
        SignatureStatus: 'Valid'
    condition: selection and not filter_legitimate
falsepositives:
    - Legitimate software installers
level: high
tags:
    - attack.defense_evasion
    - attack.t1574.002
---
title: GitHub Actions Abuse Downloading Linux Payloads
date: 2026/07/24
author: Security Arsenal
status: experimental
description: Detects processes associated with GitHub Actions runners downloading suspicious payloads from external IPs resembling the cPanel exploitation campaign.
references:
    - https://otx.alienvault.com/pulse/66a1c5b8e7b2d
logsource:
    category: process_creation
    product: linux
detection:
    selection_cli:
        CommandLine|contains:
            - 'curl'
            - 'wget'
    selection_target:
        CommandLine|contains:
            - '43.228.157.68'
            - 'api/dl'
    condition: all of selection_*
falsepositives:
    - Legitimate CI/CD workflows pulling known artifacts
level: critical
tags:
    - attack.initial_access
    - attack.t1190
---
title: Hades Implant or Go-Based Backdoor Network Traffic
date: 2026/07/24
author: Security Arsenal
status: experimental
description: Identifies suspicious network connections to known Hades C2 infrastructure or high-entropy Go binaries connecting to non-standard ports.
references:
    - https://otx.alienvault.com/pulse/66a1c5b8e7b2d
logsource:
    category: network_connection
    product: windows
detection:
    selection_ip:
        DestinationIp|contains:
            - '43.246.208.207'
    selection_port:
        DestinationPort:
            - 80
            - 443
            - 8080
    condition: selection_ip
falsepositives:
    - Low
level: high
tags:
    - attack.command_and_control
    - attack.t1071

KQL (Microsoft Sentinel)

KQL — Microsoft Sentinel / Defender
// Hunt for Mistic and Hades related IPs and Hashes
let IoC_IPs = pack_array("43.246.208.207", "43.228.157.68");
let IoC_Hashes = pack_array("3f797a639bc855bc6d5471f327924b62d10900ddec49b970eca6604142bbb4be", "818aea6143282b352fdfdc0f3ebf77a36e54eb3befb5cad1a355a99ab97c6aa7");
DeviceNetworkEvents
| where RemoteIP in (IoC_IPs) or RemoteUrl has_any("authorized-logins.net", "mueleer.com", "grande-luna.top")
| project Timestamp, DeviceName, InitiatingProcessAccountName, RemoteIP, RemoteUrl, RemotePort
| union (
    DeviceProcessEvents
    | where SHA256 in (IoC_Hashes) or ProcessCommandLine has_any("curl", "wget") and ProcessCommandLine has "api/dl"
    | project Timestamp, DeviceName, AccountName, FolderPath, SHA256, ProcessCommandLine
)

PowerShell Hunt Script

PowerShell
# IOC Hunt for Mistic, Hades, and Supply Chain Malware
# Checks for file presence and Hosts file modifications

$MaliciousHashes = @(
    "3f797a639bc855bc6d5471f327924b62d10900ddec49b970eca6604142bbb4be",
    "818aea6143282b352fdfdc0f3ebf77a36e54eb3befb5cad1a355a99ab97c6aa7",
    "881ce5cb124c4d2e814783724cc1388f6a1cbf6eee274c3f3366e77ba3503ad7"
)

$MaliciousDomains = @(
    "authorized-logins.net",
    "mueleer.com",
    "grande-luna.top",
    "laosji.net"
)

Write-Host "[+] Scanning for Malicious Files..." -ForegroundColor Cyan
$DrivesToScan = @("C:\")

foreach ($Hash in $MaliciousHashes) {
    $Result = Get-ChildItem -Path $DrivesToScan -Recurse -ErrorAction SilentlyContinue | 
              Get-FileHash -Algorithm SHA256 -ErrorAction SilentlyContinue | 
              Where-Object { $_.Hash -eq $Hash }
    
    if ($Result) {
        Write-Host "[!] MALICIOUS FILE FOUND: " $Result.Path -ForegroundColor Red
    }
}

Write-Host "[+] Checking Hosts File for C2 Domains..." -ForegroundColor Cyan
$HostsPath = "$env:SystemRoot\System32\drivers\etc\hosts"
if (Test-Path $HostsPath) {
    $HostsContent = Get-Content $HostsPath
    foreach ($Domain in $MaliciousDomains) {
        if ($HostsContent -match $Domain) {
            Write-Host "[!] SUSPICIOUS ENTRY IN HOSTS FILE: $Domain" -ForegroundColor Red
        }
    }
}

Write-Host "[+] Scan Complete." -ForegroundColor Green

Response Priorities

  • Immediate:

    • Block all listed domains and IP addresses at the firewall and proxy level.
    • Scan endpoints for the provided SHA256 hashes (Mistic, Hades, ClawHavoc samples).
    • Patch CVE-2026-41940 (cPanel/WHM), CVE-2026-43500 (Hadoop), and CVE-2021-4034 (Polkit) immediately on internet-facing systems.
  • 24 Hours:

    • Audit GitHub Actions repositories for unauthorized workflow modifications or secret leaks (checking for outbound connections to 43.228.157.68).
    • Initiate identity verification and credential reset for accounts with access to Salesforce or Gong environments (referencing the LastPass/Klue incident).
    • Investigate systems showing signs of sideloading (DLL loads from Temp/AppData).
  • 1 Week:

    • Restrict the use of third-party AI agent marketplaces (ClawHub) and implement strict code review for any imported "skills" or agents.
    • Harden CI/CD pipelines by disabling public repository access to Actions and requiring pinning of specific Action versions.
    • Conduct a comprehensive review of OAuth tokens granted to third-party vendors (SaaS supply chain audit).

Related Resources

Security Arsenal Incident Response Managed SOC & MDR Services AlertMonitor Threat Detection From The Dark Side Intel Hub

darkwebotx-pulsedarkweb-credentialsmistic-backdoorai-agent-threatgithub-actionscredential-harvestingsupply-chain

Is your security operations ready?

Get a free SOC assessment or see how AlertMonitor cuts through alert noise with automated triage.