ARTEX Attacks: Threat Actor Activity, Detection, and Response

Summary
A synthesis of incident and vendor findings describes ARTEX’s agent-harness design, reported threat-actor use, rapid attack timelines, surviving forks, observed behaviors, and detection and response guidance; some campaign outcomes remain unverified.
Key points
- ARTEX is an agent harness that coordinates a connected model, tools, and concurrent workers; its scope settings do not technically prevent network requests outside the authorized scope.
- Cloudflare described an ecosystem of more than 12,000 ARTEX-related repositories and mapped 204 public forks with code changes; these are not counts of threat actors or victims.
- Reports describe use of ARTEX or related frameworks in operations linked to groups including Storm-0940 and VAPOR PANDA; shared tooling does not establish that the groups are connected.
- Microsoft reported Storm-0940 attempted sign-ins against more than 2,250 customer tenants; a separate password spray produced more than 151,000 failed logons and no successful authentications.
- Researchers observed behaviors including rapid credential testing, cloud enumeration, unusual API requests, and bursts of activity with pauses; individual indicators may change or generate false positives.
- The article recommends validating and tuning behavioral detection rules, correlating identity, application, cloud, and network logs, and investigating whether attempts reached sensitive data.
- Recommended defenses include revoking exposed credentials, reviewing application identity permissions, securing public endpoints, and testing high-confidence automated containment.
Article Details
- Attack Vectors
- Threat actors used ARTEX and related agents to probe multiple targets and pursue attack paths concurrently.
- Reported campaigns exploited exposed applications through SQL injection and login bypasses; Storm-3358 was reported to exploit file-upload and SSRF flaws.
- An intrusion at DIVD involved exploitation of two previously unknown flaws in Zammad; the article does not specify their technical types.
- Threat actors used exposed or compromised credentials, including application identities and AWS credentials, to access secrets and cloud data.
- Microsoft reported a five-day password-spraying campaign that generated more than 151,000 failed logons and no successful authentications.
- Wiz observed repeated unsigned-JWT attempts against SSO routes; Microsoft also reported more than 200 application sign-ins in under four minutes using four credentials.
- Reported activity included cloud enumeration, attempts to access S3 data, and reading logs with compromised keys.
- Arctic Wolf associated AntiFox-linked intrusion activity with IIS webshells and stolen AWS credentials.
- Unit 42 observed agents registering accounts in a production application and reading about 50 MB of API data.
- Amazon campaign records described deepfake attempts against two crypto exchanges’ identity checks; the records said the exchanges rejected the attempts.
- CERT-In associated Crafty Falcon with unauthorized transfers through victims’ payment APIs, but did not identify which tool was used in each incident.
- Defensive Notes
- Inventory internet-facing applications and assign clear owners for maintenance, patching, and retirement; restrict management interfaces to approved access paths.
- Rotate or revoke exposed credentials, review their historical use, and investigate whether compromised identities accessed secrets stores or other credentials.
- Review application identities’ sign-ins, credentials, permissions, and subsequent data access; remove unnecessary grants and investigate unexpected changes.
- Fix authentication, authorization, and transaction-control weaknesses, including SQL injection, unsafe file uploads, SSRF, and IDOR; validate SAML and JWT signatures.
- Restrict access to internal documentation, source maps, debug output, and configuration secrets; keep detailed errors in internal logs.
- Correlate application, identity, cloud, and network logs to determine whether attempts succeeded and what accounts, systems, or data were accessed.
- Use automated containment only for well-defined, high-confidence detections; test playbooks and recovery procedures, preserve evidence, and verify that actions actually stop access.
- Enforce testing-agent target restrictions with network controls and narrowly scoped credentials; use decoys in controlled environments.
- Validate and tune detection rules against local logs, account for authorized testing, and correlate behavioral signals rather than blocking solely on changeable tool markers or aging IPs.
MITRE ATT&CK
T1078 · Valid AccountsMicrosoft reported that Storm-0940 used compromised application identities to attempt sign-ins across customer tenants and gained privileged access at some organizations.T1110.003 · Password SprayingMicrosoft reported a five-day password-spraying campaign that produced more than 151,000 failed logons and no successful authentications.T1190 · Exploit Public-Facing ApplicationReported campaigns exploited exposed applications through SQL injection, login bypasses, file-upload flaws, and SSRF; the DIVD breach involved two previously unknown Zammad flaws.T1505.003 · Web ShellArctic Wolf associated AntiFox-linked intrusion activity with IIS webshells; the article also describes detection coverage for webshell upload followed by execution.T1526 · Cloud Service DiscoveryThe reported activity includes cloud enumeration and attempts to discover or access cloud resources using compromised keys.T1530 · Data from Cloud StorageArctic Wolf reported successful cloud actions involving access to S3 data during intrusions associated with AntiFox infrastructure.T1552 · Unsecured CredentialsThe article describes searches for exposed secrets and campaigns involving exposed or stolen credentials, including AWS credentials and LLM API keys.T1567.004 · Exfiltration Over WebhookThe article describes unexpected outbound traffic sending output to rotating webhook.site URLs.T1595 · Active ScanningThe article reports scanning from infrastructure associated with Storm-3358 and describes agents probing multiple targets.
People
Autumn-27The article identifies Autumn-27 as the ARTEX creator’s GitHub handle; the creator said the tool was intended for authorized testing, denied involvement in the attacks, and ended public releases. Their real name was not verified.Hendrik Adrian (RickGPT)Developed the ARTEX detection pack, which includes behavioral detection rules and supporting validation and tuning material.
Threat Actors
Crafty FalconCERT-In associated Crafty Falcon with attacks on lenders, fintechs, and crypto exchanges, including unauthorized transfers through victims’ payment APIs. The advisory did not identify which tool was used in each incident.Onyx SleetMicrosoft named Onyx Sleet among the North Korean groups using ARTEX and said people continued to direct the operations.Pearl SleetMicrosoft named Pearl Sleet among the North Korean groups using ARTEX and said people continued to direct the operations.Storm-0940Microsoft reported that Storm-0940 used ARTEX after compromising application identities behind ConnectWise’s cloud backup product, attempted sign-ins at more than 2,250 customer tenants, and established privileged persistence at many organizations.Storm-3358Microsoft described Storm-3358 as a Cambodia-based, financially motivated cluster targeting Southeast Asian financial firms and exploiting file-upload and SSRF flaws. Cisco Talos reported scanning from associated infrastructure, which corroborates activity but not every element of Microsoft’s attribution.TAG-217Recorded Future names the group TAG-217. Zscaler reported SQL injection and login-bypass activity and data theft affecting about 67,000 people; CrowdStrike assessed with moderate confidence that the group was Chinese-speaking and financially motivated.VAPOR PANDACrowdStrike reported that VAPOR PANDA began a global campaign on October 17, targeted more than 200 organizations, and used agents directed by a human operator, including to find machines without an endpoint sensor.
Vendors
Amazon Web Servicesincidents and customer telemetry from the research of Unit 42, Google GTIG/Mandiant, CrowdStrike, Wiz, AWS, Microsoft, Datadog and Check Point, plus guidance from various CERTs and the FBI. Together, their researchAnthropicNovember 2025: the earlier warning. Anthropic disclosed that a Chinese state-sponsored group had jailbroken Claude Code and used it against roughly 30 targets. This predates ARTEX and concerns a separate operation.ConnectWiseand official warnings. VAPOR PANDA begins its global campaign. Microsoft connects Storm-0940 to the ConnectWise compromise. A researcher’s ARTEX test accesses a WorkOS internal tool without authorization. SeoulMicrosoftand customer telemetry from the research of Unit 42, Google GTIG/Mandiant, CrowdStrike, Wiz, AWS, Microsoft, Datadog and Check Point, plus guidance from various CERTs and the FBI. Together, their research givesThreatBookThreatBook’s October 16 announcement says ARTEX began as a personal project by RedAgents’ product lead. It says the open-source edition won Baidu’s Agent+ challenge on September 3, and acknowledges that it lacked
Products
Tools
AntiFoxCloudflare’s incident responders have encountered both ARTEX and ARTEX-derived frameworks. AntiFox is one such fork: its maintainer documents its origin as an ARTEX v0.1.1 fork. Arctic Wolf associated AntiFox interfacesARTEXIf you’re anything like me, you’ve probably come across “ARTEX” a few hundred times over the past month. You’ve probably also felt the uptick and acceleration in attacks. So what is it, at a granular level? Are weARTEX PROon September 3, and acknowledges that it lacked built-in security controls. RedAgents includes Artex Pro, a proprietary enterprise edition, alongside other testing agents. ThreatBook advertises authorizationCairntransfers through victims’ own payment APIs. The advisory lists ARTEX, PentAGI, CyberStrikeAI and Cairn, a separate AI framework that searches for a sequence of actions to reach a goal. It does not identify theCyberStrikeAIunauthorized transfers through victims’ own payment APIs. The advisory lists ARTEX, PentAGI, CyberStrikeAI and Cairn, a separate AI framework that searches for a sequence of actions to reach a goal. It doesHexstrikeHexStrike - an MCP server that lets AI agents run pentesting tools.PentAGIexchanges, including unauthorized transfers through victims’ own payment APIs. The advisory lists ARTEX, PentAGI, CyberStrikeAI and Cairn, a separate AI framework that searches for a sequence of actions to reach a
Countries
CambodiaSeoul police announce seven arrests. Microsoft reports use by two North Korean groups and a Cambodia-based crime cluster. Google, Check Point, Censys, Unit 42, Cloudflare, Amazon and Cisco Talos publishChinaChina state-aligned, according to Microsoft. After compromising application identities behind ConnectWise’s cloud backup product, it used ARTEX to attempt sign-ins at more than 2,250 customer tenants. Microsoft saysNetherlandsAugust; Microsoft dates Storm-0940’s first use to September. On September 21, a TA breaches DIVD, the Dutch vulnerability disclosure nonprofit, through two previously unknown flaws in Zammad. DIVD later confirmsNorth Koreainternal tool without authorization. Seoul police announce seven arrests. Microsoft reports use by two North Korean groups and a Cambodia-based crime cluster. Google, Check Point, Censys, Unit 42, Cloudflare, AmazonUnited Statestracked activity across at least 14 countries in 11 days. Targets included foreign ministries and U.S. pharmaceutical, semiconductor, defense and nuclear organizations. CrowdStrike confirmed activity after
Industries
Defense14 countries in 11 days. Targets included foreign ministries and U.S. pharmaceutical, semiconductor, defense and nuclear organizations. CrowdStrike confirmed activity after initial compromise at nine organizations.GovernmentGovernment and CERTs: FBI FLASH-20261113-001 · CERT-EU · CERT-In CIAD-2026-0058 · Korea’s regulator count, via Herald Business (Korean) and Akamai
