A few days ago, GTIG documented that ARTEX, AntiFox, and PentAGI are running live offensive operations — autonomous AI frameworks conducting credential theft, lateral movement, and infrastructure compromise in real environments. The FBI had already issued a FLASH alert on November 13 naming healthcare as a target sector. If you read the November 18 post on the GTIG findings, you know what these frameworks are and what they're capable of. This post is the follow-on: a hunting guide, built from @m4rio_eth's nine-step methodology published November 16, that tells you how to actually find them in your environment.
The gap between "we know these frameworks exist" and "we can detect them in our logs" is where most healthcare organizations are sitting right now. This post is an attempt to close that gap.
Why These Three Frameworks, Why Now
ARTEX (linked to Storm-0940/Midnight Blizzard infrastructure), AntiFox, and PentAGI are not academic curiosities. ARTEX is the framework GTIG tied directly to live credential-theft campaigns. AntiFox is an IIS-targeting framework that leaves a recognizable process chain on Windows servers. PentAGI is an open-source pentest agent that ships with identifiable default configuration — including a TLS certificate with pentagi.local in the subject. All three have known, stable IOCs and behavioral signatures. That's the hunting opportunity.Healthcare environments are specifically at risk for two structural reasons. First, IIS is ubiquitous — it runs patient portals, revenue cycle interfaces, and legacy EHR integrations. AntiFox specifically targets IIS web servers. Second, healthcare organizations are heavy users of RMM (Remote Monitoring and Management) tools like ConnectWise Automate, which ARTEX has been observed staging from. If your SOC doesn't have visibility into ConnectWise application installs, that's a gap worth closing this week.
The Nine-Step Hunting Methodology
The m4rio_eth guide structures detection across three layers: network/infrastructure, host telemetry, and framework-specific fingerprints. What follows is a healthcare-contextualized walk-through of each detection surface.Step 1: Threat Actor Infrastructure — ARTEX
Start with the known-bad IPs. ARTEX's command-and-control infrastructure as documented by GTIG includes three primary nodes:- 144.172.64[.]60
- 144.172.64[.]99
- 144.172.64[.]128
Query these across your firewall logs, proxy logs, and DNS resolvers. If you use a SIEM, build a persistent alert. For healthcare organizations running EDR, check for outbound connections to this /24 range from any clinical system or identity infrastructure. ARTEX routes command output through rotating webhook.site endpoints, so if you see frequent outbound traffic to webhook.site from servers that have no business reason to reach it, that's a secondary indicator.
Step 2: AntiFox Network IOCs
AntiFox's known C2 infrastructure includes two IPs: 47.79.4[.]33 and 8.208.101[.]210. Same playbook: firewall logs, proxy, DNS. Healthcare-specific note: if these appear in traffic from your patient-facing web tier or any IIS-hosted application, escalate immediately. AntiFox's operational pattern is to establish persistence on IIS, then reach out for further tasking.Step 3: ConnectWise Application ID Blocklist — ARTEX
This is a step that most healthcare SOCs are not doing, and it's one of the highest-value detections on the list. ARTEX has been observed using ConnectWise Automate to establish its initial foothold. Three specific application IDs are associated with ARTEX deployments:- 24205314-2b89-4e07-a2a4-579e898a5640
- 0de14c77-b9b0-4137-9360-34878a929827
- 506e08f9-6e9a-4da4-942b-a099a03644ae
If your organization manages ConnectWise Automate agents (or allows vendors to install them), these IDs should be in your blocklist. This requires coordination with your IT ops team and your ConnectWise admin. If you don't know which ConnectWise application IDs are authorized in your environment, that inventory gap is worth flagging to your RMM team now.
Step 4: AntiFox IIS File Artifact Patterns
AntiFox leaves a recognizable file pattern on compromised IIS servers. When it stages on a system, it drops files matching three naming patterns in the web root or accessible directories:- cfg*.txt — configuration exfiltration staging
- wc*.txt and el*.txt — data collection artifacts
- pwn*.aspx — webshell with predictable naming
For healthcare environments running IIS, a scheduled file integrity check against these patterns in your web root directories is a low-cost detection control. If you have FIM (File Integrity Monitoring) on your IIS servers, add these patterns to your alert configuration. The pwn*.aspx pattern should be an immediate escalation trigger — there's no legitimate reason for a file with that prefix in a web directory.
Step 5: AntiFox Process Chain on IIS Servers
AntiFox's process execution pattern is identifiable in Windows event logs and EDR telemetry. The chain to look for is:w3wp.exe (IIS worker process) spawning any of:
- powershell.exe
- cmd.exe
- whoami.exe
In a healthy IIS environment, the worker process should not be spawning command shells or running reconnaissance commands. If you're running an EDR with process tree visibility (CrowdStrike, Defender for Endpoint, SentinelOne), create a detection rule for w3wp.exe as a parent of powershell.exe or cmd.exe. This is a high-fidelity signal — false positive rate is low for most healthcare IIS deployments.
For healthcare organizations without EDR on all IIS servers, Windows Security event logs (Event ID 4688 with process creation auditing enabled) can surface this chain, though at lower fidelity than EDR.
Step 6: ARTEX Application Fingerprinting
If you have web application scanning or active recon capability, ARTEX has two reliable fingerprints. First, the HTML page title of ARTEX's C2 panel is consistent across deployments — if you're scanning infrastructure for exposed management interfaces and you see this title, you've likely found ARTEX staging infrastructure. Second, ARTEX responses include a distinctive HTTP header: X-Artex-Auth. If you're doing passive TLS/HTTP inspection on outbound traffic and see this header, that's a high-confidence indicator of ARTEX communication.For healthcare security teams doing vendor risk assessments or scanning third-party-managed systems, these fingerprints are worth including in your scan templates.
Step 7: AntiFox Application Fingerprinting
AntiFox has two parallel fingerprints. The browser console of AntiFox's management panel contains a distinctive title string. Additionally, AntiFox API responses include service: antifox in their JSON structure. As with ARTEX, if you're doing any active scanning of infrastructure or reviewing HTTP response content from internal hosts, these strings are reliable AntiFox indicators.Step 8: PentAGI Default Configuration Exposure
PentAGI is an open-source agentic pentest framework, which means it ships with defaults that many deployments never change. The two most useful for detection: PentAGI exposes a metadata endpoint at /api/v1/info that returns version and configuration information. The default TLS certificate uses pentagi.local as the common name and subject alternative name.If you're scanning your network — or if you're reviewing certificates in your internal PKI or proxy logs — the pentagi.local certificate is a strong signal that someone has deployed PentAGI inside your perimeter. This matters because PentAGI has legitimate uses as a pentest tool, so the question is whether your red team or a third-party assessor authorized that deployment.
Step 9: Correlate Across Layers
The highest-confidence detections come from correlation, not individual IOC hits. If you see outbound traffic to 144.172.64[.]99 from a server that also has an unauthorized ConnectWise application installed, that's a compound signal. If you see w3wp.exe spawning powershell.exe on an IIS server that also has cfg*.txt files in the web root, that's not a coincidence.Build a correlation search in your SIEM that looks for two or more of these indicators on the same host within a 24-48 hour window. That's your escalation trigger.
What This Means for Healthcare
Healthcare organizations face a specific version of this threat that's worth naming directly.IIS Is Your Highest-Priority Detection Surface
The AntiFox detection steps — file patterns, process chain, application fingerprints — are all IIS-specific. Healthcare is IIS-heavy by legacy. Patient portals, revenue cycle applications, and integration engines built on Microsoft technology stacks are common in mid-size and large health systems. If your security team doesn't have clear visibility into process execution on IIS servers, AntiFox's foothold will look like normal IIS noise until it isn't.The first action from this post should be: inventory your IIS servers, verify EDR coverage, and add w3wp.exe child process detection if it doesn't already exist.
ConnectWise and RMM Blindspots
ARTEX's use of ConnectWise as an initial access vector directly targets a common healthcare operational pattern. Many health systems rely on RMM tools for managed service providers, IT outsourcing, and biomedical device support. Those same tools are attack surfaces. If your SOC doesn't receive telemetry from your ConnectWise environment — or if you don't have a process for authorizing new ConnectWise application IDs — that's a gap that this threat actively exploits.This isn't a new problem. It's the same RMM abuse pattern seen in the Change Healthcare intrusion pathway. The difference is that ARTEX automates the follow-on activity once it has that foothold.
Log Coverage Gaps Are the Real Risk
The nine-step methodology assumes you have the logs to query. Many healthcare organizations don't — or they have them but they're not being analyzed. ARTEX routes output through webhook.site because it blends into normal outbound web traffic. AntiFox uses process chains that require process creation auditing to be enabled in Windows event logs. PentAGI's fingerprints are only visible if you're doing TLS inspection or active scanning.The practical question for a healthcare security team is: for each of these nine detection steps, do you have the data source to run the query? Build that inventory. Where you have gaps, those are your prioritized telemetry investments. Entra sign-in logs, AWS CloudTrail, IIS access logs, and Windows process creation events (4688) are the core set.
Automated Frameworks Lower the Skill Floor
The HIPAA Security Rule's risk analysis requirements don't distinguish between a human attacker and an AI-driven one. What matters is that the threat to ePHI is real and that the safeguards are in place. What the GTIG report and this hunting guide together establish is that the barrier to running a sophisticated, multi-stage intrusion has dropped significantly. Healthcare organizations that treated AI-driven offensive frameworks as a future concern as of six months ago are now looking at present-tense threat activity.Running the Hunt
If you're doing a one-time hunt rather than standing up persistent detections, here's a reasonable sequence for a healthcare security team with a half-day time investment.Start with the network IOCs — firewall and proxy query for all three ARTEX IPs and both AntiFox IPs across 90 days of logs. If you're in a healthcare environment with historical log retention for compliance, you have the data. Flag anything that hits and pivot to the host.
Move to ConnectWise application IDs — pull the authorized application list from your RMM admin and diff it against the three ARTEX IDs. If you can't get that list, that's the finding.
Then run file pattern searches on IIS web roots. This is a filesystem query that can be scripted with PowerShell in under ten minutes:
Get-ChildItem -Path "C:\inetpub\wwwroot" -Recurse -Include "cfg*.txt","wc*.txt","el*.txt","pwn*.aspx"
Finally, check your EDR console for process tree alerts involving w3wp.exe as a parent process. If you don't have EDR on all IIS servers, pull Windows Security event 4688 for w3wp.exe parent within the same timeframe as your network IOC query.
Document what you found and what you couldn't query. The gap list is as useful as the hits.
The Bigger Picture
The GTIG report established that agentic offensive frameworks have crossed into active, operational use. The m4rio_eth hunting guide is the practitioner response — here are the specific, queryable signals these frameworks leave behind. That combination of threat documentation and detection methodology is exactly what healthcare security teams need to move from awareness to action.The frameworks themselves will evolve. ARTEX's C2 IPs will rotate. AntiFox's file naming patterns will be updated when attackers know defenders are hunting them. PentAGI's defaults will get changed by operators who read detection guides. That's the nature of this work. The value of running this hunt now is establishing a baseline — what does your environment look like when you run these queries against historical data? If you find nothing, you've confirmed coverage. If you find hits, you have an incident.
Either outcome is better than not knowing.
This is an entry in the AI Security series. For related coverage, see Autonomous AI Offensive Frameworks: GTIG Documents Machine-Speed Threats to Healthcare.
Key Links
- @m4rio_eth: Hunting the hunters — ARTEX, AntiFox and PentAGI (X/Twitter) [Free]
- Google GTIG: Agentic Threat Activity Crosses a Significant Threshold [Free]
- PentAGI GitHub Repository [Free]
- CISA Cybersecurity Advisories (Storm-0940/ARTEX context) [Free]
- Microsoft: Event 4688 — Process Creation Auditing Reference [Free]