1. Executive summary
Microsoft Threat Intelligence reports that a financially motivated cluster it tracks as Storm-3358 began in October 2026 targeting financial organisations in Southeast Asia and a smaller number of large financial organisations globally, using the agentic AI offensive framework ARTEX alongside conventional automated tooling. The actor scans internet-facing systems, exploits server-side request forgery and anonymous file-upload flaws, and installs open-source remote access trojans (WISP, vshell and AdaptixC2) for persistence and lateral movement. No CVE is associated with the exploited flaws in public reporting, no CVSS score or CISA KEV entry exists for them, and no fixed release is named. For EMEA financial services the direct exposure is to internet-facing web applications that accept anonymous uploads or make server-side requests, and to the cloud identities reachable once such an application is compromised; the campaign as reported has not named an EMEA victim.
2. Regulatory framing
No specific DORA/NIS2 article is directly engaged by this item.
3. Technical analysis & attack chain
Microsoft's 6 November 2026 report describes an intrusion set active since October 2026. The confirmed steps Microsoft gives are:
- ARTEX agents scan systems connected to the public internet.
- The actor exploits security flaws in web applications, specifically anonymous file upload and server-side request forgery (SSRF).
- Open-source remote access trojans are installed: WISP, vshell and AdaptixC2.
- ARTEX agents use utilities including netcat and curl to download the RATs onto compromised systems.
- The RATs provide persistence and lateral movement.
- Across recent ARTEX-related attacks more broadly, Microsoft reports agents using web shells for persistence on compromised web servers, harvesting credentials from public sources and from compromised systems, and using compromised cloud identities for persistence and data exfiltration.
The exploited flaws are described only by class. No CVE identifier, affected product, affected or fixed version, or patch is named in Microsoft's report or in the Cybersecurity Dive coverage. NVD has not scored them because no CVE is attached. The specific request payloads used against the SSRF and upload flaws are not published.
ARTEX is an agentic offensive framework, not a single implant. Microsoft states that by default it uses Nmap, curl, wget, Python HTTP clients, Norma SDK WebFetch, and headless Chromium through Playwright. Its agents can also perform brute-force password attacks, password spraying, API enumeration and abuse, and database dumping. The operational consequence is speed: Microsoft says agentic tools can use compromised credentials and systems at machine speed to move laterally, escalate privileges and exfiltrate data. Microsoft notes that actors routinely route this traffic through proxies and VPNs to change source IP, so source-IP blocking is not a durable control.
Detection of the agentic phase rests on traffic shape rather than a fixed signature. Microsoft describes frameworks sending a wide variety of requests in a brief time, retrying variations on the same technique, or moving to next-stage actions at machine speed, and notes that the default tooling may emit default user agent strings. Microsoft is explicit that blocking on those characteristics will not stop an agent, which will change them quickly; the value is in alerting.
Attribution and confidence. Microsoft attributes the activity to Storm-3358 and describes it as a Cambodia-based intrusion cluster that primarily targets financial organisations in Southeast Asia plus a smaller number of large financial organisations globally. Storm-3358 has no MITRE ATT&CK profile, so the attribution is unconfirmed and rests on a single vendor's tracking. The victim count, the identities of victims and any EMEA presence are not stated in public reporting. The claim that the actor is financially motivated is Microsoft's characterisation; no source states a motive beyond that label.
Separate events. Cybersecurity Dive places this campaign alongside other ARTEX activity: attacks on South Korean banks and dozens of Japanese companies; abuse of ConnectWise Cloud Backup against Microsoft 365 tenants linked to a China-based actor tracked as Storm-0940; a CrowdStrike-linked global espionage campaign by an actor tracked as Vapor Panda against 200 organisations; an Indian CERT-In report on ARTEX, Cairn and PentAGI use against fintechs, digital lenders and non-banking financial companies by an actor tracked as Crafty Falcon; and Microsoft reporting of ARTEX used to scan GitHub for exposed credentials. These are distinct actors and events. None of their indicators, victims or techniques should be read into Storm-3358, and none of the four named actors has a MITRE ATT&CK profile, so all of those attributions are unconfirmed.
4. Mitigation & containment
P1, within 24 hours:
- Inventory every internet-facing web application that accepts file uploads and confirm whether anonymous upload is possible. Where it is not a documented business requirement, disable anonymous upload or require authentication before the upload handler is reached.
- Review web application logs for server-side requests originating from the application itself to internal addresses or to cloud metadata endpoints. Microsoft does not publish the SSRF payloads, so the control is to establish whether the application can be made to issue such requests at all.
- Hunt for netcat and curl spawning from web server or application server processes, which is the download step Microsoft describes for the RATs. Where application servers have no legitimate need for either binary, alert on and then block their execution from those process trees.
- Search web-accessible directories on internet-facing servers for web shells, which Microsoft reports ARTEX-related actors using for persistence on compromised web servers.
P2, within 72 hours:
- Alert on the default tooling Microsoft names for ARTEX: Nmap, curl, wget, Python HTTP clients, Norma SDK WebFetch, and headless Chromium through Playwright. Treat default user agent strings from these tools as an alerting signal, not a block, because Microsoft states an agent will change them quickly.
- Baseline request volume per client against internet-facing applications and alert on bursts of varied requests with rapid retries against the same endpoint, the pattern Microsoft attributes to agentic scanning.
- Review cloud identity activity for the persistence and exfiltration pattern Microsoft describes: credentials harvested from public sources or compromised systems, then used from cloud identities. Require MFA and conditional access on identities that can reach data stores from application infrastructure.
- Deploy decoys, honeypots, tarpits and fake-success surfaces in front of real assets. Microsoft reports that in its testing these reduced attack traffic reaching real assets and produced high-confidence signal, and that fake wins can cause an autonomous agent to stop on a false success.
P3, within 7 days:
- Add the brute-force, password-spraying, API enumeration and database-dumping behaviours Microsoft lists for ARTEX agents to authentication and API monitoring thresholds.
- Confirm that no fixed release exists to apply for the exploited SSRF and upload flaws: none is named in public reporting. Compensating controls (authentication in front of upload handlers, egress restriction from application servers, network segmentation between web tier and data tier) are the available remediation until a vendor fix is identified.
5. Indicators of compromise
No atomic indicators of compromise are published in Microsoft's report or in the Cybersecurity Dive coverage. Microsoft states it is providing IOCs in its blog, but no hashes, domains, IP addresses or file paths appear in the material available here. The RAT names (WISP, vshell, AdaptixC2) and the utility names (netcat, curl, Nmap, wget, Playwright, Norma SDK WebFetch) are tool identities, not pivotable indicators, and are covered as behaviours below.
Behavioural indicators
| behaviour | where to observe | confidence |
|---|---|---|
| Wide variety of requests to internet-facing applications in a brief time, with retries and variations on the same technique | Web application and WAF logs | High, stated by Microsoft |
| Requests carrying default user agent strings from Nmap, curl, wget, Python HTTP clients, Norma SDK WebFetch or headless Chromium via Playwright | Web server and proxy logs | Medium, Microsoft notes agents change these quickly |
| netcat or curl invoked to download a file onto a compromised host | EDR process telemetry on web and application servers | High, stated by Microsoft |
| Web shell written to a compromised web server | File integrity monitoring on web-accessible directories | High, stated by Microsoft for ARTEX-related attacks |
| Credentials harvested from public sources or compromised systems then used from cloud identities for persistence and exfiltration | Cloud identity and authentication logs | High, stated by Microsoft |
| Brute-force, password spraying, API enumeration and database dumping | Authentication logs, API gateway logs, database audit logs | High, stated by Microsoft for ARTEX agents |
6. Detection
The sources name tools and behaviours but publish no strings, command lines, filenames, paths, registry keys or mutexes from the malicious artefacts themselves. A YARA rule built on tool names or actor names would detect reporting about the threat, not the threat, so none is emitted. The Sigma rule below keys on the download behaviour Microsoft describes (netcat or curl fetching a file onto a host) and is a behavioural hunt, not a signature.
title: Possible ARTEX RAT download via curl or netcat
id: 7c1e4b2a-3f6d-4a91-9c2e-5b8d0f1a6e34
status: experimental
description: Detects curl or netcat used to download a file to disk, the step Microsoft reports ARTEX agents using to pull WISP, vshell and AdaptixC2 onto compromised systems.
author: Adverse Trace
date: 2026-11-09
references:
- https://www.microsoft.com/en-us/security/blog/2026/11/06/artex-abuse-expands-as-cybercriminals-target-financial-sector/
logsource:
category: process_creation
product: windows
detection:
selection_curl:
Image|endswith:
- '\curl.exe'
CommandLine|contains:
- ' -o '
- ' --output '
- ' -O '
selection_netcat:
Image|endswith:
- '\nc.exe'
- '\ncat.exe'
- '\netcat.exe'
condition: selection_curl or selection_netcat
falsepositives:
- Administrative scripts and software installers that use curl or netcat legitimately
level: medium
Tune the netcat branch to your estate: on Linux application servers the same behaviour appears as nc, ncat or netcat in process telemetry, and Microsoft's report does not restrict the download step to Windows.
7. Sources
- Microsoft Threat Intelligence, "ARTEX abuse expands as cybercriminals target financial sector", https://www.microsoft.com/en-us/security/blog/2026/11/06/artex-abuse-expands-as-cybercriminals-target-financial-sector/, 2026-11-06
- Cybersecurity Dive, "Financially motivated threat cluster uses AI-enabled tools against banking sector", https://www.cybersecuritydive.com/news/financially-motivated-threat-ai-tools-banking-artex/874116/, 2026-11-09
8. Adverse Trace position
Adverse Trace assesses this as a medium-severity item for EMEA financial services. The actor is real and active against financial targets, the exploited flaw classes (SSRF and anonymous file upload) are common in customer-facing banking applications, and the RATs named provide persistence and lateral movement. Against that, no EMEA victim is named, no CVE or CVSS score exists to prioritise patching, no fixed release is available, and the entire technical picture rests on a single vendor report (Microsoft) plus one trade-press write-up of it. The Storm-3358 attribution is single-sourced and unconfirmed, and the actor has no MITRE ATT&CK profile. We will monitor for a CVE assignment or vendor advisory covering the SSRF and upload flaws, for Microsoft's published IOCs, and for any reporting that places a victim in EMEA financial services; if any of those appear we will reissue this advisory with a revised severity and, where a fix exists, a patch-specific P1. Clients running internet-facing applications with anonymous upload or server-side request capability should treat the P1 items in section 4 as due now rather than waiting for a CVE.
Published via PulseTrace — Adverse Trace threat intelligence.