Back to Intelligence

BUGSTER AI-Assisted Infostealer Ops + 'Skynet' LLMjacking Agent: OTX Pulse Analysis — Enterprise Detection Pack

SA
Security Arsenal Team
November 20, 2026
9 min read

Five fresh OTX pulses paint a converging picture: financially motivated threat actors are industrializing credential theft by chaining AI tooling, disposable phishing infrastructure, and abuse of trusted enterprise software. This briefing synthesizes live pulse data and ships detection logic you can deploy today.


Threat Summary

The pulses collectively reveal a maturing cybercrime ecosystem where AI is the force multiplier:

  1. BUGSTER (AI-Assisted Intrusion Ops) — A low-skill, financially motivated actor leveraged a jailbroken Hermes Agent framework backed by DeepSeek V4 to run reconnaissance, generate exploitation scripts, and exfiltrate 5,200+ personal records from an education institution in India, with telecommunications also targeted. AI collapsed the skill barrier: the actor executed multi-stage operations that would previously have required an experienced intrusion team.

  2. One Phishing Kit, 29 Fake Banks — A single templated phishing kit has spawned 29 unique fake-bank domains targeting US and Canadian financial customers since late September 2026. Identical artifacts across all sites (same tagline, routing number, support phone) confirm a common operator. Objective: credential and PII harvesting at scale using disposable domains.

  3. 'Skynet' Autonomous LLMjacking Agent — An autonomous offensive agent scans the internet for exposed LLM endpoints and hijacks them to run attacks. It carries 74 offensive tools covering recon, web framework exploitation (VS Code, Django, Jupyter, Laravel, Supabase, Next.js, Flask/Werkzeug), database theft via Prisma, credential harvesting, and cryptomining via XMRig.

  4. Phishing → LLMjacking on GCP — Actors use sponsored search ads pushing fake Google login pages to steal privileged GCP credentials, then manipulate organizational policies to disable guardrails, create service accounts with Gemini API access, and invite attacker-controlled accounts — hijacking cloud billing for LLM compute (cross-tenant exploitation).

  5. PowerPanel Business (PPB) Intrusion Vector — Actors abused CyberPower UPS management software on a Windows domain controller, uploading malicious .cmd files via the web console on TCP 3052. Commands executed through ppbed.exe run as SYSTEM, enabling Active Directory extraction (ntds.dit theft) via DLL side-loading and MSIL.Killav!gen1 for AV tampering.

Common thread: every pulse terminates in credential theft, identity abuse, or monetized compute/data theft. The attack chain of the moment is: phish or expose → steal privileged credentials → disable guardrails → monetize (PII sale, LLM compute, AD kingdom keys).


Threat Actor / Malware Profile

BUGSTER (AI-Augmented Intrusion Actor)

  • Distribution: AI-generated scripts delivered post-recon; exploits CVE-2026-109618 and CVE-2026-109619 against education/telecom targets.
  • Payload behavior: AI-authored exfiltration tooling; bulk PII theft (5,200+ records confirmed).
  • Skill signature: Heavy reliance on Hermes Agent + DeepSeek V4 output; expect sloppy but functional code, verbose comments, and generic variable naming in recovered scripts.

Skynet LLMjacking Agent + XMRig

  • Distribution: Internet-wide scanning for exposed LLM endpoints (Ollama, unauthenticated API gateways).
  • Payload behavior: 74-tool offensive arsenal; framework-specific exploitation modules; Prisma-based database theft; credential harvesting; drops XMRig for monetization.
  • C2 communication: Mining pool infrastructure — mine.c3pool.com:13333, auto.c3pool.org:80, auto.skypool.xyz:6666. Stratum protocol over TCP.
  • Persistence: Compromised LLM endpoints themselves become persistent attack platforms (the victim pays the compute bill).

MSIL.Killav!gen1 + PPB Abuse

  • Distribution: Manual upload of .cmd files via PowerPanel Business web console (TCP 3052).
  • Payload behavior: ppbed.exe executes attacker commands as SYSTEM; DLL side-loading for stealth; ntds.dit extraction from the domain controller.
  • Anti-analysis: Killav family terminates security tooling; execution rides a legitimate signed service, evading application allowlisting.

GCP Phishing → LLMjacking Cluster

  • Distribution: Sponsored search ads (malvertising) → lookalike login domains (groosl.com, gboogl.com, gcoalg.com, etc.).
  • Payload behavior: Org-policy manipulation to disable constraints, rogue service account creation with Gemini API scopes, external account invitation for cross-tenant persistence.

IOC Analysis

Indicator TypeCountOperationalization
Domains (phishing/lookalike)37+DNS sinkhole + proxy block. Lookalike patterns (g*o*l.com, *bnk*, *finance*) are regex-friendly for proactive detection.
Hostnames/URLs (mining C2)5Block at egress; alert on Stratum ports 13333/6666/80 to pool infrastructure.
FileHash-SHA256 (XMRig, Killav, PPB tooling)18Load into EDR block lists; sweep via hash hunt across all endpoints.
CVEs2Emergency patch verification for CVE-2026-109618 / CVE-2026-109619 on internet-facing education/telecom assets.

Tooling: OTX indicators ingest cleanly via TAXII/STIX into Sentinel, Splunk ES, or MISP. Decode and enrich domains with passive DNS (SecurityTrails, VirusTotal) to catch the next 29 domains from the banking kit — identical page artifacts mean passive-DNS clustering on shared registrars/name servers will surface unlisted infrastructure. Hash indicators should be cross-referenced against internal EDR telemetry retroactively for 90 days.


Detection Engineering

YAML
---
title: PowerPanel Business Service Spawning Suspicious Child Processes
id: 9f3a1c2e-7b41-4d55-9a2c-ppb3052sys01
status: experimental
description: Detects ppbed.exe (CyberPower PowerPanel Business) spawning command shells or scripting engines, consistent with OTX pulse reporting SYSTEM-level command execution via uploaded .cmd files leading to AD extraction.
author: Security Arsenal Threat Intel
references:
  - https://www.security.com/blog-post/power-management-software-attack-vector
date: 2026/11/21
logsource:
  category: process_creation
  product: windows
detection:
  selection_parent:
    ParentImage|endswith: '\ppbed.exe'
  selection_child:
    Image|endswith:
      - '\cmd.exe'
      - '\powershell.exe'
      - '\pwsh.exe'
      - '\wscript.exe'
      - '\cscript.exe'
      - '\rundll32.exe'
      - '\regsvr32.exe'
      - '\ntdsutil.exe'
      - '\vssadmin.exe'
  condition: selection_parent and selection_child
falsepositives:
  - Legitimate PPB power-event scripts configured by administrators
level: high
tags:
  - attack.privilege_escalation
  - attack.t1059
  - attack.t1574.002
---
title: XMRig Mining Pool Network Connections (Skynet LLMjacking)
id: 2c8d4f1a-6e92-4ab3-b7d1-xmrig1333302
status: experimental
description: Detects outbound connections to known c3pool/skypool mining infrastructure associated with XMRig deployment by the autonomous Skynet LLMjacking agent.
author: Security Arsenal Threat Intel
references:
  - https://www.security.com/blog-post/ai-llmjacking-skynet
date: 2026/11/21
logsource:
  category: network_connection
  product: windows
detection:
  selection_domain:
    DestinationHostname|contains:
      - 'c3pool.com'
      - 'c3pool.org'
      - 'skypool.xyz'
  selection_port:
    DestinationPort:
      - 13333
      - 6666
  condition: selection_domain or (selection_port and 1 of selection_initiator*)
  selection_initiator:
    Initiated: 'true'
falsepositives:
  - Legitimate cryptomining (policy violation, still alert-worthy)
level: high
tags:
  - attack.impact
  - attack.t1496
  - attack.t1071
---
title: Fake Banking / GCP Lookalike Phishing Domain Resolution
id: 7a1b9c3d-2f48-4e66-a5b9-phishkit2903
status: experimental
description: Detects DNS resolution of known fake-bank phishing kit domains and GCP lookalike credential-harvesting domains reported in OTX pulses.
author: Security Arsenal Threat Intel
date: 2026/11/21
logsource:
  category: dns
  product: windows
detection:
  selection_exact:
    query|contains:
      - 'brayfinance.online'
      - 'firstcp.live'
      - 'securecapitalcorporationlimited.com'
      - 'qtbnk.com'
      - 'norexgo.com'
      - '1stggbinc.cc'
      - 'oakmontfb.com'
      - 'allyunitybn.com'
      - 'groosl.com'
      - 'groonl.com'
      - 'gboogl.com'
      - 'gbrokl.space'
      - 'gbrokl.xyz'
      - 'gcoalg.com'
      - 'gcslom.com'
      - 'gpoosl.com'
  condition: selection_exact
falsepositives:
  - Threat research / sandbox detonation
level: critical
tags:
  - attack.initial_access
  - attack.t1566.002
  - attack.t1078
KQL — Microsoft Sentinel / Defender
// Security Arsenal — OTX Pulse Hunt: PPB abuse, XMRig C2, phishing kit domains
let phishingDomains = dynamic(["brayfinance.online","bank.firstcp.live","securecapitalcorporationlimited.com","qtbnk.com","norexgo.com","1stggbinc.cc","oakmontfb.com","allyunitybn.com","groosl.com","groonl.com","gboogl.com","gbrokl.space","gbrokl.xyz","gcoalg.com","gcslom.com","gpoosl.com"]);
let miningC2 = dynamic(["mine.c3pool.com","auto.c3pool.org","auto.skypool.xyz"]);
let iocHashes = dynamic(["b720a5e7c3a5b1760d15a8833a782f4fe7fc5c62b967fe57f607e72745ec0ccf","fcde1323acceb46fd88f2276c010bb98839dccce03dc512c56d9529428ceec68","04206431c83d1d81726e56fffb5166c0a7f138eb366e5aa918dee54af1ef22b5","12e716f48875befa978ff4a78a0b99a18b1809fc5bcc34c761158875da611aa8","1949894c26e03d6f15ed7f4393a89dbd52f8acc62959d3d5e4f695d6a19c6d9c","22b8f2776f1c941d8240ee4243da537147f8d2ee4d859a68038335da7b0d7688","2776cfcdd46264043b823adaf37e0c85d0efd3b46d5fdbb72219535061d44b86","3d73d5342a11dce35081e70a4ea75d39ad65bc6d480337bd33c1d2e0ba4090ee","5b2285f788eec4d17bf18d85795d457f40f7598bbab2697919f5810f54a9c3fe","76670eaadda9806c9aefa91c2ee4f2a33dcf8124574721a6ff072f19e7c03514"]);
let lookback = 14d;
union isfuzzy=true
    (DeviceNetworkEvents
    | where TimeGenerated > ago(lookback)
    | where RemoteUrl has_any (phishingDomains) or RemoteUrl has_any (miningC2)
    | project TimeGenerated, DeviceName, InitiatingProcessFileName, RemoteUrl, RemotePort, ActionType, Hit="NetworkIOC"),
    (DeviceProcessEvents
    | where TimeGenerated > ago(lookback)
    | where SHA256 has_any (iocHashes)
    | project TimeGenerated, DeviceName, FileName, FolderPath, SHA256, ProcessCommandLine, Hit="FileHashIOC"),
    (DeviceProcessEvents
    | where TimeGenerated > ago(lookback)
    | where InitiatingProcessFileName =~ "ppbed.exe"
    | where FileName in~ ("cmd.exe","powershell.exe","rundll32.exe","ntdsutil.exe","vssadmin.exe")
    | project TimeGenerated, DeviceName, FileName, ProcessCommandLine, InitiatingProcessFileName, Hit="PPBAbuse"),
    (SecurityEvent
    | where TimeGenerated > ago(lookback)
    | where EventID == 4688
    | where ParentProcessName endswith "ppbed.exe"
    | project TimeGenerated, Computer, NewProcessName, CommandLine, ParentProcessName, Hit="PPBAbuse-SecEvent")
| sort by TimeGenerated desc
PowerShell
# Security Arsenal — OTX Pulse IOC Hunt (PowerShell)
# Sweeps for PPB abuse artifacts, XMRig hashes/connections, and phishing DNS cache hits
$ErrorActionPreference = 'SilentlyContinue'
$report = @()

# 1. Known malicious SHA256 hashes from OTX pulses
$hashes = @(
 'b720a5e7c3a5b1760d15a8833a782f4fe7fc5c62b967fe57f607e72745ec0ccf',
 'fcde1323acceb46fd88f2276c010bb98839dccce03dc512c56d9529428ceec68',
 '04206431c83d1d81726e56fffb5166c0a7f138eb366e5aa918dee54af1ef22b5',
 '12e716f48875befa978ff4a78a0b99a18b1809fc5bcc34c761158875da611aa8',
 '1949894c26e03d6f15ed7f4393a89dbd52f8acc62959d3d5e4f695d6a19c6d9c',
 '22b8f2776f1c941d8240ee4243da537147f8d2ee4d859a68038335da7b0d7688',
 '2776cfcdd46264043b823adaf37e0c85d0efd3b46d5fdbb72219535061d44b86',
 '3d73d5342a11dce35081e70a4ea75d39ad65bc6d480337bd33c1d2e0ba4090ee',
 '5b2285f788eec4d17bf18d85795d457f40f7598bbab2697919f5810f54a9c3fe',
 '76670eaadda9806c9aefa91c2ee4f2a33dcf8124574721a6ff072f19e7c03514'
)

# 2. Hash sweep of common staging paths
$paths = @("$env:TEMP","$env:ProgramData","C:\Windows\Temp","C:\Program Files\CyberPower","C:\Program Files (x86)\CyberPower")
foreach ($p in $paths) {
  Get-ChildItem -Path $p -Recurse -File -ErrorAction SilentlyContinue | ForEach-Object {
    $h = (Get-FileHash $_.FullName -Algorithm SHA256).Hash.ToLower()
    if ($hashes -contains $h) {
      $report += [pscustomobject]@{Hit='MaliciousHash'; Detail=$_.FullName; SHA256=$h}
    }
  }
}

# 3. PPB service + suspicious script artifacts (power-event .cmd uploads)
Get-Service | Where-Object { $_.Name -match 'ppbe|PowerPanel' } | ForEach-Object {
  $report += [pscustomobject]@{Hit='PPBServicePresent'; Detail=$_.Name; SHA256=''}
}
Get-ChildItem "C:\Program Files*\CyberPower" -Recurse -Include *.cmd,*.bat,*.ps1 -ErrorAction SilentlyContinue | ForEach-Object {
  $report += [pscustomobject]@{Hit='PPBScriptArtifact'; Detail=$_.FullName; SHA256=''}
}

# 4. Active/historical connections to mining C2
$c2 = @('mine.c3pool.com','auto.c3pool.org','auto.skypool.xyz')
Get-NetTCPConnection -State Established | Where-Object {
  $_.RemotePort -in @(13333,6666) -or ($c2 | ForEach-Object { $_ }) -contains (Resolve-DnsName $_.RemoteAddress -ErrorAction SilentlyContinue).NameHost
} | ForEach-Object {
  $report += [pscustomobject]@{Hit='MiningC2Connection'; Detail="$($_.RemoteAddress):$($_.RemotePort)"; SHA256=''}
}

# 5. DNS cache check for phishing kit domains
$phish = @('brayfinance.online','qtbnk.com','norexgo.com','oakmontfb.com','allyunitybn.com','groosl.com','groonl.com','gboogl.com','gbrokl.xyz','gcoalg.com','gcslom.com','gpoosl.com')
Get-DnsClientCache | Where-Object { $e=$_.Entry; $phish | Where-Object { $e -like "*$_*" } } | ForEach-Object {
  $report += [pscustomobject]@{Hit='PhishingDNSCache'; Detail=$_.Entry; SHA256=''}
}

# 6. Suspicious scheduled tasks (Killav persistence check)
Get-ScheduledTask | Where-Object { $_.Actions.Execute -match 'cmd|powershell|rundll32' -and $_.TaskPath -notlike '\Microsoft*' } | ForEach-Object {
  $report += [pscustomobject]@{Hit='SuspiciousScheduledTask'; Detail="$($_.TaskPath)$($_.TaskName) -> $($_.Actions.Execute)"; SHA256=''}
}

$report | Format-Table -AutoSize
$report | Export-Csv -Path ".\otx_hunt_$(Get-Date -Format yyyyMMdd_HHmm).csv" -NoTypeInformation
Write-Host "[+] Hunt complete. $($report.Count) findings." -ForegroundColor Cyan

Response Priorities

Immediate (0–4h)

  • Block all listed phishing domains, mining C2 hostnames/ports (13333, 6666), and file hashes at DNS, proxy, and EDR layers.
  • Hunt for ppbed.exe child processes on all systems running CyberPower PowerPanel Business — prioritize domain controllers. Restrict or firewall TCP 3052 to management subnets only.
  • Sweep for XMRig presence (hash hunt + Stratum egress connections), especially on AI/LLM hosting infrastructure.

24 Hours

  • Any host that resolved a fake-bank or GCP lookalike domain: force password reset + revoke active sessions/refresh tokens for the associated user. Assume credential compromise.
  • GCP environments: audit IAM for newly created service accounts with Gemini API scopes, external member additions, and recent orgpolicy constraint modifications. Rotate all privileged credentials.
  • If PPB artifacts are found on a DC: treat as full AD compromise — ntds.dit extraction means krbtgt double-reset and enterprise-wide credential rotation.

1 Week

  • Audit internet exposure of LLM endpoints (Ollama, API gateways); require authentication and network segmentation. Deploy egress filtering blocking mining pools.
  • Patch CVE-2026-109618 and CVE-2026-109619 on exposed education/telecom-facing assets.
  • Implement brand-lookalike domain monitoring (regex-based passive DNS alerting) for your organization's name and for Google/banking lures targeting staff.
  • Enforce phishing-resistant MFA (FIDO2) for all privileged cloud console access — sponsored-ad phishing defeats SMS/TOTP.

Related Resources

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

Is your security operations ready?

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