The three OTX pulses converge on one operational theme: adversaries are industrializing access by abusing exposed AI/developer infrastructure, consumer connected-TV firmware, and low-skill AI-assisted intrusion workflows. BUGSTER represents the democratized end of the spectrum: a financially motivated actor using jailbroken Hermes Agent workflows with DeepSeek V4 to generate reconnaissance, scripts, and exploitation support against education and telecommunications targets in India, with two fresh CVEs flagged for exploit opportunity. The Skynet pulse shows autonomous LLMjacking: internet-scanning agents locate exposed LLM endpoints and weaponize 74 offensive tools against VS Code, Django, Jupyter, Laravel, Supabase, Next.js, Flask/Werkzeug, Prisma-backed data stores, credential material, and cryptomining deployment using XMRig. BADBOX 3.0 is the scaled monetization layer: backdoored low-cost consumer devices and TV boxes linked to TXCZ generate roughly 40 billion bid requests per day and more than 4% of connected-TV traffic through invisible ad rendering and residential-proxy behavior.
Collectively, the attack chain is: discover exposed services and weak devices, coerce AI or prebuilt modules into execution, establish outbound C2 or proxy monetization, then cash out through data theft, credential harvesting, cryptomining, or ad-fraud bid volume. Enterprise risk is not confined to classic endpoints. SOC scope must include developer workstations, cloud AI gateways, Jupyter/Flask/Laravel edge apps, Android/CTV supply-chain exposure, and egress to mining, proxy, and ad-fraud infrastructure.
Threat Actor / Malware Profile
BUGSTER is assessed as financially motivated and technically dependent on AI-generated tradecraft. Distribution is likely spear-phishing, exposed application exploitation, and opportunistic CVE probing rather than a single custom loader. Payload behavior centers on AI-assisted reconnaissance, script generation, record extraction, and follow-on extortion or resale. Persistence is probably lightweight: scheduled tasks, web shells, stolen tokens, or abused legitimate remote access. Anti-analysis appears indirect: jailbroken AI prompts, disposable scripts, and living-off-the-land commands reduce consistent binary artifacts.
Skynet LLMjacking agent hijacks exposed LLM endpoints as free compute and as an attack orchestration layer. Its tool coverage maps to discovery, web-framework exploitation, database theft through Prisma/ORM paths, credential harvesting, and XMRig deployment. C2 and monetization infrastructure in the pulse includes mining pool endpoints such as mine.c3pool.com, auto.c3pool.org, and auto.skypool.xyz over ports 13333, 80, and 6666. Persistence commonly lands as cron/systemd units, container escape remnants, rogue Python/Node services, or miner watchdog processes. Anti-analysis includes process name masquerading, CPU throttling, log deletion, and use of legitimate pool protocols.
BADBOX 3.0 / TXCZ-linked modules are supply-chain and firmware-adjacent. Infected low-cost Android/CTV devices receive backdoors that pull fraud modules, render invisible video ads, impersonate legitimate CTV apps, and route traffic through residential-proxy infrastructure. Named infrastructure includes pxyedge.com, file.drgdj.top, gh.drgdj.top, and IPv4 indicators 107.151.239.17, 165.154.199.119, 165.154.239.242, 165.154.254.58, and 199.190.45.202. Persistence is strongest at device image, OTA update, launcher, or vendor service level. Anti-analysis relies on invisible ad views, traffic blending with CTV ad tech, proxy rotation, and modular payloads delivered after initial compromise.
IOC Analysis
The indicator set is network-heavy, which is useful because BADBOX and Skynet monetization requires reliable egress. Prioritize domain, hostname, URL, and IPv4 blocking at DNS resolver, secure web gateway, egress firewall, and cloud NAT layers. The SHA256 hashes b720a5e7c3a5b1760d15a8833a782f4fe7fc5c62b967fe57f607e72745ec0ccf and fcde1323acceb46fd88f2276c010bb98839dccce03dc512c56d9529428ceec68 should be queried in EDR, malware detonation, threat intel platforms, and object storage for miner or agent payloads. CVE-2026-109618 and CVE-2026-109619 require immediate asset inventory matching for education, telecom, and internet-facing development stacks even before public exploit maturity is confirmed.
Operationalization model: convert domains and hostnames to DNS sinkhole and RPZ entries; add URLs to proxy categories for cryptomining, proxy avoidance, and malware C2; push IPv4s to edge deny lists with short TTL review; import hashes into EDR blocklists and sandbox retro-hunt; map CVEs to external attack surface management and WAF virtual patches. Tooling that decodes and enriches these indicators includes MISP/OpenCTI for TIP normalization, CrowdStrike/Defender/SentinelOne retro-hunt for hashes, Zeek/Suricata for DNS and JA3/JA4 anomalies, Splunk/Sentinel for egress baselining, urlscan.io and VirusTotal for URL detonation, and Shodan/Censys for exposed LLM, Jupyter, Laravel debug, and VS Code server surfaces.
Detection Engineering
---
title: BADBOX 3.0 or Skynet Egress to Ad Fraud Proxy and Cryptomining Infrastructure
id: 8e5f2c1a-otx-2026-0001
status: experimental
description: Detects DNS, proxy, or network egress to OTX indicators tied to BADBOX 3.0 ad-fraud C2 and Skynet LLMjacking/XMRig pool infrastructure.
author: Security Arsenal Detection Engineering
date: 2026/11/21
references:
- https://otx.alienvault.com/
logsource:
category: network_connection
product: zeek
detection:
selection_dns:
query|contains:
- pxyedge.com
- file.drgdj.top
- gh.drgdj.top
- mine.c3pool.com
- auto.c3pool.org
- auto.skypool.xyz
selection_ip:
DestinationIp|in:
- 107.151.239.17
- 165.154.199.119
- 165.154.239.242
- 165.154.254.58
- 199.190.45.202
selection_url:
c-uri|contains:
- http://mine.c3pool.com:13333
- http://auto.c3pool.org:80
- http://auto.skypool.xyz:6666
condition: 1 of selection_*
falsepositives:
- Rare legitimate research or sandbox detonation traffic
level: high
tags:
- attack.command_and_control
- attack.impact
- attack.t1496
---
title: Suspicious XMRig Miner Execution After LLM or Developer Tool Exposure
id: 8e5f2c1a-otx-2026-0002
status: experimental
description: Identifies process and command-line patterns consistent with XMRig deployment following Skynet-style LLMjacking of exposed AI, Jupyter, VS Code, Flask, Laravel, or Next.js services.
author: Security Arsenal Detection Engineering
date: 2026/11/21
references:
- https://otx.alienvault.com/
logsource:
category: process_creation
product: windows
detection:
selection_img:
Image|endswith:
- \xmrig.exe
- \miner.exe
- \python.exe
- \node.exe
- \powershell.exe
selection_cli:
CommandLine|contains:
- c3pool
- skypool
- --donate-level
- --coin=monero
- stratum+tcp
- 13333
- 6666
selection_parent:
ParentImage|endswith:
- \code.exe
- \jupyter-notebook.exe
- \python.exe
- \php.exe
- \node.exe
condition: selection_img and selection_cli and selection_parent
falsepositives:
- Authorized benchmarking or security research with explicit change record
level: critical
tags:
- attack.execution
- attack.t1059
- attack.impact
- attack.t1496
---
title: Linux Persistence and Miner Watchdog From Exposed AI Edge Services
id: 8e5f2c1a-otx-2026-0003
status: experimental
description: Detects Linux host persistence and miner watchdog behavior associated with Skynet LLMjacking and BADBOX-like modular backdoors on servers, containers, and embedded devices.
author: Security Arsenal Detection Engineering
date: 2026/11/21
references:
- https://otx.alienvault.com/
logsource:
category: process_creation
product: linux
detection:
selection_cron_systemd:
CommandLine|contains:
- crontab -
- systemctl enable
- /etc/systemd/system/
- /var/tmp/
- /dev/shm/
selection_miner:
CommandLine|contains:
- xmrig
- c3pool
- skypool
- monero
- stratum
- 13333
- 6666
selection_net_tool:
CommandLine|contains:
- curl http
- wget http
- base64 -d
- chmod +x
condition: selection_cron_systemd and selection_miner and selection_net_tool
falsepositives:
- Developer automation pulling signed packages from approved repositories
level: high
tags:
- attack.persistence
- attack.t1053
- attack.t1543
- attack.resource_development
let bad_iocs = dynamic([
"pxyedge.com","file.drgdj.top","gh.drgdj.top","mine.c3pool.com","auto.c3pool.org","auto.skypool.xyz",
"107.151.239.17","165.154.199.119","165.154.239.242","165.154.254.58","199.190.45.202"
]);
let bad_hashes = dynamic([
"b720a5e7c3a5b1760d15a8833a782f4fe7fc5c62b967fe57f607e72745ec0ccf",
"fcde1323acceb46fd88f2276c010bb98839dccce03dc512c56d9529428ceec68"
]);
union isfuzzy=true
(DeviceNetworkEvents
| where TimeGenerated > ago(7d)
| where RemoteUrl has_any (bad_iocs) or RemoteIP in (bad_iocs)
| project TimeGenerated, DeviceName, InitiatingProcessFileName, InitiatingProcessCommandLine, RemoteIP, RemotePort, RemoteUrl, ActionType),
(DeviceProcessEvents
| where TimeGenerated > ago(7d)
| where SHA256 in (bad_hashes)
or ProcessCommandLine has_any ("xmrig","c3pool","skypool","stratum","13333","6666","pxyedge","drgdj")
or FileName in~ ("xmrig.exe","miner.exe","xmrig")
| project TimeGenerated, DeviceName, FileName, ProcessCommandLine, SHA256, FolderPath, InitiatingProcessFileName),
(DeviceFileEvents
| where TimeGenerated > ago(7d)
| where SHA256 in (bad_hashes) or FolderPath has_any ("/var/tmp/","/dev/shm/","\\Temp\\","\\AppData\\Local\\Temp\\")
| project TimeGenerated, DeviceName, FileName, FolderPath, SHA256, InitiatingProcessFileName)
| sort by TimeGenerated desc
$ErrorActionPreference = 'SilentlyContinue'
$iocDomains = @('pxyedge.com','file.drgdj.top','gh.drgdj.top','mine.c3pool.com','auto.c3pool.org','auto.skypool.xyz')
$iocIps = @('107.151.239.17','165.154.199.119','165.154.239.242','165.154.254.58','199.190.45.202')
$iocHash = @('b720a5e7c3a5b1760d15a8833a782f4fe7fc5c62b967fe57f607e72745ec0ccf','fcde1323acceb46fd88f2276c010bb98839dccce03dc512c56d9529428ceec68')
Write-Host '[+] Active network IOC matches'
Get-NetTCPConnection | Where-Object { $_.RemoteAddress -in $iocIps -or $_.RemotePort -in 13333,6666 } | Select-Object LocalAddress,LocalPort,RemoteAddress,RemotePort,State,OwningProcess
Write-Host '[+] Suspicious miner/ad-fraud processes'
Get-Process | Where-Object { $_.ProcessName -match 'xmrig|miner|monero|python|node|powershell|php' } | Select-Object Id,ProcessName,Path,CommandLine
Write-Host '[+] DNS cache hits for OTX domains'
Get-DnsClientCache | Where-Object { $n=$_.Name; $iocDomains | Where-Object { $n -like ('*'+$_) } } | Select-Object Name,Type,Data
Write-Host '[+] Persistence review: tasks, services, startup, Run keys'
Get-ScheduledTask | Where-Object { $_.TaskPath -notlike '\Microsoft*' -and ($_.Actions.Execute -match 'powershell|cmd|python|node|xmrig' -or $_.Actions.Arguments -match 'c3pool|skypool|pxyedge|drgdj') } | Select-Object TaskName,TaskPath,State
Get-CimInstance Win32_Service | Where-Object { $_.PathName -match 'Temp|AppData|var/tmp|dev/shm|xmrig|c3pool|skypool|pxyedge|drgdj' } | Select-Object Name,State,StartMode,PathName
Get-ItemProperty 'HKLM:\Software\Microsoft\Windows\CurrentVersion\Run','HKCU:\Software\Microsoft\Windows\CurrentVersion\Run' | Format-List
Write-Host '[+] Hash sweep in high-risk writable paths'
$paths = @("$env:TEMP","$env:LOCALAPPDATA\Temp",'C:\ProgramData','/var/tmp','/dev/shm')
foreach ($p in $paths) { if (Test-Path $p) { Get-ChildItem $p -Recurse -File -ErrorAction SilentlyContinue | ForEach-Object { $h=(Get-FileHash $_.FullName -Algorithm SHA256).Hash.ToLower(); if ($iocHash -contains $h) { Write-Host "[!] IOC hash hit: $($_.FullName) $h" } } } }
Response Priorities
Immediate: Block listed domains, hostnames, URLs, IPs, and SHA256 hashes at DNS, SWG, EDR, firewall, cloud egress, and mail gateway. Hunt for XMRig command lines, c3pool/skypool egress, pxyedge/drgdj resolution, exposed LLM endpoint logs, and unexpected python/node/php children of web or IDE processes. Isolate CTV/IoT VLANs and deny direct internet from unmanaged streaming devices.
24 hours: Because Skynet explicitly covers credential harvesting and BUGSTER operations include record exfiltration, force identity verification for admin, developer, cloud, SaaS, VPN, and service accounts with recent anomalous sign-ins. Rotate secrets found in .env, Jupyter config, VS Code settings, Laravel env files, Supabase service keys, Prisma schemas, CI variables, and browser credential stores. Review token grants to AI coding assistants and revoke unexplained OAuth consent.
1 week: Harden architecture based on vector: require auth and private networking for LLM gateways, Jupyter, VS Code Server, admin panels, and database tooling; disable debug modes for Laravel/Django/Flask; add WAF virtual patches for CVE-2026-109618 and CVE-2026-109619 exposure; segment CTV/IoT from enterprise identity; enforce egress allowlists for servers; implement application control for miner binaries; require SBOM and firmware provenance review for low-cost Android/CTV procurement; and add detections for invisible ad rendering, residential-proxy beacons, and abnormal bid-request telemetry.
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.