Storm-3358 ARTEX attacks reached financial-sector organizations in October 2026, Microsoft Threat Intelligence said in a 6 November 2026 blog post. Storm-3358 is the name Microsoft gives to a Cambodia-based, financially motivated intrusion cluster, and ARTEX is the agentic AI offensive framework the group used.
Microsoft says Storm-3358 mainly targets financial organizations in Southeast Asia, plus a smaller number of large financial organizations worldwide. This is a different actor from two other ARTEX uses Microsoft has covered this year: a nation-state group behind a supply-chain compromise. The other was a separate campaign scanning GitHub for exposed OAuth secrets, which Microsoft’s own post does not attribute to a named actor. This post does not merge any of them with Storm-3358.
Why the Storm-3358 ARTEX Campaign Matters for Financial Defenders
ARTEX is an open-source AI pentest framework. Microsoft says Storm-3358 pairs it with ordinary automated tools to scan systems connected to the public internet. The actor then exploits flaws, including anonymous file upload and server-side request forgery (SSRF), to get a foothold.
Microsoft frames this inside a wider pattern: threat actors using agentic AI pentest frameworks, ARTEX among them, to speed up every stage of an attack. Agentic tools can act at machine speed. They try variations on one technique, or move to the next step, without a human operator in the loop.
Three Microsoft blog posts now cover a distinct actor or activity abusing ARTEX. This one covers Storm-3358 and the financial sector. An earlier post covered a nation-state supply-chain campaign, and another covered agentic tools searching GitHub for exposed OAuth secrets.
Microsoft treats each of the three as a separate finding, with its own actor or activity pattern. Brandefense keeps that same separation in its own coverage, and this post does not combine any of their details with Storm-3358’s campaign.
How Storm-3358 Gets In and Moves Around
Once an ARTEX agent finds an opening, Microsoft says it uses utilities such as netcat and curl to pull down remote access trojans (RATs). Storm-3358 has used three open-source RATs for persistence and lateral movement: WISP, vshell and AdaptixC2.
- Internet-facing appInternetSSRF or anonymous upload exploited
- Web shell plantedEdgenetcat and curl fetch a RAT
- WISP, vshell or AdaptixC2InternalCredential and API abuse
- Compromised cloud identityInternalLateral movement
- Data collectedData
Microsoft describes a wider set of techniques across recent ARTEX-related attacks, not limited to Storm-3358 alone. These include web shells left on compromised web servers and credentials pulled from public sources and compromised systems. Compromised cloud identities are then used to stay in an environment and move data out.
ARTEX agents can also run brute-force password attacks, password spraying, API enumeration and abuse, and database dumping, Microsoft says. Agentic tools can use stolen credentials and compromised systems at machine speed. They move laterally, raise their privileges and pull data out faster than a human-run intrusion typically would.
Microsoft also notes that threat actors regularly route this traffic through proxies and VPNs. That changes the source IP address and defeats static IP blocklists. It is one reason Microsoft’s own guidance leans on behavioral detection rather than a fixed list of bad addresses.
Why Financially Motivated Actors Target Banks With Agentic Tools
Microsoft’s own description ties Storm-3358 to financial motivation rather than espionage or disruption. A financially motivated group gains directly from access to payment systems, customer data or fraud paths inside a bank’s network. That is a different goal from the nation-state campaign Microsoft covered separately.
Southeast Asia has a concentration of mid-sized financial organizations, alongside a smaller set of large global ones Microsoft also names as targets. Brandefense’s own reading is that attackers often test new tooling against a regional set of similar targets before widening their reach. Microsoft’s post does not state this as the group’s plan, so this is our own pattern reading, not a confirmed motive.
The practical effect for a defender is the same regardless of motive. An internet-facing application with an exploitable SSRF or upload flaw is a target, whether the actor behind it wants money, espionage access or disruption. The controls in the next two sections apply across all three.
What to Check First: Upload Forms and SSRF Exposure
Financial-sector security teams should start with their own internet-facing attack surface. Any form that accepts a file upload without authentication is a candidate for abuse. So is any service that lets a server fetch a URL on a user’s behalf, the classic SSRF pattern.
Brandefense recommends treating both exposure types as a priority inventory item this quarter. List every anonymous upload endpoint and every SSRF-capable integration on internet-facing systems, then confirm each one is authenticated, rate-limited or removed.
Review third-party and vendor-facing upload portals too, not only customer-facing ones. Microsoft’s description of the entry vector does not name specific products, so Brandefense’s advice here is our own reading of the pattern, not a claim about any named application.
Detecting Storm-3358 ARTEX Activity
Microsoft says agentic AI frameworks leave behavioral traces even when their output varies. A scan run by ARTEX sends many requests in a short time. It then retries variations on one technique, or moves straight to the next stage, all at machine speed.
By default, ARTEX uses existing penetration-testing tools, including Nmap, curl, wget, Python HTTP clients, the Norma SDK’s WebFetch function, and headless Chromium automated through Playwright. Microsoft notes these tools can carry distinctive traffic traits, such as default user-agent strings, though an attacker can change them quickly once blocked.
Microsoft’s own detection list for this campaign spans several tactics: process execution, data collection, credential access, defense evasion, discovery, command-and-control and lateral movement. Example alerts in that list include a process launched in the background and enumeration of files with sensitive data. A possible reverse shell is another, and together they sketch the shape of an intrusion rather than one single tell.
What Storm-3358 ARTEX Targets Should Do Now
Microsoft points financial-sector defenders toward identity risk signals first: sign-in risk detections and conditional access policies that require multifactor authentication when a session looks risky. Storm-3358 traffic should register as activity from known threat-actor infrastructure where that detection is enabled.
Endpoint telemetry is the second priority. Microsoft’s own detection list gives defenders concrete alert categories to review. That matters more than a single signature for a framework built to change its own indicators between runs.
Brandefense’s own reading of this campaign is that financial-sector teams should treat the entry vector, not the RAT choice, as the control point. WISP, vshell and AdaptixC2 are open-source tools a different actor could swap out tomorrow. Anonymous upload and SSRF exposure are fixed properties of an application until someone closes them.
For the underlying tool, see Brandefense’s profile of the ARTEX AI pentest framework. For the entry techniques this campaign relies on, see exploiting public-facing applications and remote access trojans. Brandefense has also covered two other campaigns that reused ARTEX: a supply-chain compromise against ConnectWise customers and a search for exposed OAuth secrets on GitHub.