Blog
/
OT
/
April 16, 2026

Inside ZionSiphon: Darktrace’s Analysis of OT Malware Targeting Israeli Water Systems

Darktrace analysis reveals ZionSiphon, an OT‑focused malware targeting Israeli water treatment and desalination systems. The malware combines privilege escalation, persistence, USB propagation, and ICS scanning with sabotage capabilities aimed at chlorine and pressure controls, highlighting growing experimentation with politically motivated critical ‑infrastructure attacks against industrial operational technologies globally.
Inside the SOC
Darktrace cyber analysts are world-class experts in threat intelligence, threat hunting and incident response, and provide 24/7 SOC support to thousands of Darktrace customers around the globe. Inside the SOC is exclusively authored by these experts, providing analysis of cyber incidents and threat trends, based on real-world experience in the field.
Written by
Calum Hall
Technical Content Researcher
zionsiphonDefault blog image
16
Apr 2026

What is ZionSiphon?

Darktrace recently analyzed a malware sample, which identifies itself as ZionSiphon. This sample combines several familiar host-based capabilities, including privilege escalation, persistence, and removable-media propagation, with targeting logic themed around water treatment and desalination environments.

This blog details Darktrace’s investigation of ZionSiphon, focusing on how the malware identifies targets, establishes persistence, attempts to tamper with local configuration files, and scans for Operational Technology (OT)-relevant services on the local subnet. The analysis also assesses what the code suggests about the threat actor’s intended objectives and highlights where the implementation appears incomplete.

Function “ZionSiphon()” used by the malware author.
Figure 1: Function “ZionSiphon()” used by the malware author.

ZionSiphon targets and motivations

Israel-Focused Targeting and Messaging‍

The clearest indicators of intent in this sample are its hardcoded Israel-focused targeting checks and the strong political messaging found in some strings in the malware’s binary.

In the class initializer, the malware defines a set of IPv4 ranges, including “2.52.0.0-2.55.255.255”, “79.176.0.0-79.191.255.255”, and “212.150.0.0-212.150.255.255”, indicating that the author intended to restrict execution to a narrow range of addresses. All of the specified IP blocks are geographically located within Israel.

The malware obfuscates the IP ranges by encoding them in Base64.
Figure 2: The malware obfuscates the IP ranges by encoding them in Base64.

The ideological motivations behind this malware are also seemingly evident in two Base64-encoded strings embedded in the binary. The first (shown in Figure 1) is:

“Netanyahu = SW4gc3VwcG9ydCBvZiBvdXIgYnJvdGhlcnMgaW4gSXJhbiwgUGFsZXN0aW5lLCBhbmQgWWVtZW4gYWdhaW5zdCBaaW9uaXN0IGFnZ3Jlc3Npb24uIEkgYW0gIjB4SUNTIi4=“, which decodes to “In support of our brothers in Iran, Palestine, and Yemen against Zionist aggression. I am "0xICS".

The second string, “Dimona = UG9pc29uaW5nIHRoZSBwb3B1bGF0aW9uIG9mIFRlbCBBdml2IGFuZCBIYWlmYQo=“, decodes to “Poisoning the population of Tel Aviv and Haifa”.  These strings do not appear to be used by the malware for any operational purpose, but they do offer an indication of the attacker’s motivations. Dimona, referenced in the second string, is an Israeli city in the Negev desert, primarily known as the site of the Shimon Peres Negev Nuclear Research Center.

The Dimona string as it appears in the decompiled malware, with the Base64-decoded text.
Figure 3: The Dimona string as it appears in the decompiled malware, with the Base64-decoded text.

The hardcoded IP ranges and propaganda‑style text suggest politically motivated intent, with Israel appearing to be a likely target.

Water and desalination-themed targeting?

The malware also includes Israel-linked strings in its target list, including “Mekorot, “Sorek”, “Hadera”, “Ashdod”, “Palmachim”, and “Shafdan”. All of the strings correspond to components of Israel’s national water infrastructure: Mekorot is Israel’s national water company responsible for managing the country’s water system, including major desalination and wastewater projects. Sorek, Hadera, Ashdod, and Palmachim are four of Israel’s five major seawater desalination plants, each producing tens of millions of cubic meters of drinking water annually. Shafdan is the country’s central wastewater treatment and reclamation facility. Their inclusion in ZionSiphon’s targeting list suggests an interest in infrastructure linked to Israel’s water sector.

Strings in the target list, all related to Israel and water treatment.
Figure 4: Strings in the target list, all related to Israel and water treatment.

Beyond geographic targeting, the sample contains a second layer of environment-specific checks aimed at water treatment and desalination systems. In the function ”IsDamDesalinationPlant()”, the malware first inspects running process names for strings such as “DesalPLC”, “ROController”, “SchneiderRO”, “DamRO”, “ReverseOsmosis”, “WaterGenix”, “RO_Pump”, “ChlorineCtrl”, “WaterPLC”, “SeaWaterRO”, “BrineControl”, “OsmosisPLC”, “DesalMonitor”, “RO_Filter”, “ChlorineDose”, “RO_Membrane”, “DesalFlow”, “WaterTreat”, and “SalinityCtrl”. These strings are directly related to desalination, reverse osmosis, chlorine handling, and plant control components typically seen in the water treatment industry.

The filesystem checks reinforce this focus. The code looks for directories such as “C:\Program Files\Desalination”, “C:\Program Files\Schneider Electric\Desal”, “C:\Program Files\IDE Technologies”, “C:\Program Files\Water Treatment”, “C:\Program Files\RO Systems”, “C:\Program Files\DesalTech”, “C:\Program Files\Aqua Solutions”, and “C:\Program Files\Hydro Systems”, as well as files including “C:\DesalConfig.ini”, “C:\ROConfig.ini”, “C:\DesalSettings.conf”, “C:\Program Files\Desalination\system.cfg”, “C:\WaterTreatment.ini”, “C:\ChlorineControl.dat”, “C:\RO_PumpSettings.ini”, and “C:\SalinityControl.ini.”

Malware Analysis

Privilege Escalation

The “RunAsAdmin” function from the malware sample.
Figure 5: The “RunAsAdmin” function from the malware sample.


The malware’s first major action is to check whether it is running with administrative rights. The “RunAsAdmin()” function calls “IsElevated()”, which retrieves the current Windows identity and checks whether it belongs to the local Administrators group. If the process is already elevated, execution proceeds normally.

The “IsElevated” function as seen in the sample.
Figure 6: The “IsElevated” function as seen in the sample.


If not, the code waits on the named mutex and launches “powershell.exe” with the argument “Start-Process -FilePath <current executable> -Verb RunAs”, after which it waits for that process to finish and then exits. ‍

Persistence and stealth installation

Registry key creation.
Figure 7: Registry key creation.

Persistence is handled by “s1()”. This routine opens “HKCU\Software\Microsoft\Windows\CurrentVersion\Run”, retrieves the current process path, and compares it to “stealthPath”. If the current file is not already running from that location, it copies itself to the stealth path and sets the copied file’s attributes to “hidden”.

The code then creates a “Run” value named “SystemHealthCheck” pointing to the stealth path. Because “stealthPath” is built from “LocalApplicationData” and the hardcoded filename “svchost.exe”, the result is a user-level persistence mechanism that disguises the payload under a familiar Windows process name. The combination of a hidden file and a plausible-sounding autorun value suggests an intent to blend into ordinary Windows artifacts rather than relying on more complex persistence methods.

Target determination

The malware’s targeting determination is divided between “IsTargetCountry()” and “IsDamDesalinationPlant()”. The “IsTargetCountry()” function retrieves the local IPv4 address, converts it to a numeric value, and compares it against each of the hardcoded ranges stored in “ipRanges”. Only if the address falls within one of these ranges does the code move on to next string-comparison step, which ultimately determines whether the country check succeeded.

The main target validation function.
Figure 8: The main target validation function.
 The “IsTargetCountry” function.
Figure 9 : The “IsTargetCountry” function.


“IsDamDesalinationPlant()” then assesses whether the host resembles a relevant OT environment. It first scans running process names for the hardcoded strings previously mentioned, followed by checks for the presence of any of the hardcoded directories or files. The intended logic is clear: the payload activates only when both a geographic condition and an environment specific condition related to desalination or water treatment are met.

Figure. 10: An excerpt of the list of strings used in the “IsDamDesalinationPlant” function

Why this version appears dysfunctional

Although the file contains sabotage, scanning, and propagation functions, the current sample appears unable to satisfy its own target-country checking function even when the reported IP falls within the specified ranges. In the static constructor, every “ipRanges” entry is associated with the same decoded string, “Nqvbdk”, derived from “TnF2YmRr”. Later, “IsTargetCountry()” (shown in Figure 8) compares that stored value against “EncryptDecrypt("Israel", 5)”.

The “EncryptDecrypt” function
Figure 11: The “EncryptDecrypt” function

As implemented, “EncryptDecrypt("Israel", 5)” does not produce “Nqvbdk”, it produces a different string. This function seems to be a basic XOR encode/decode routine, XORing the string “Israel” with value of 5. Because the resulting output does not match “Nqvbdk” the comparison always fails, even when the host IP falls within one of the specified ranges. As a result, this build appears to consistently determine that the device is not a valid target. This behavior suggests that the version is either intentionally disabled, incorrectly configured, or left in an unfinished state. In fact, there is no XOR key that would transform “Israel” into “Nqvbdk” using this function.‍

Self-destruct function

The “SelfDestruct” function
Figure 12: The “SelfDestruct” function

If IsTargetCountry() returns false, the malware invokes “SelfDestruct()”. This routine removes the SystemHealthCheck value from “HKCU\Software\Microsoft\Windows\CurrentVersion\Run”, writes a log file to “%TEMP%\target_verify.log” containing the message “Target not matched. Operation restricted to IL ranges. Self-destruct initiated.” and creates the batch file “%TEMP%\delete.bat”. This file repeatedly attempts to delete the malware’s executable, before deleting itself.

Local configuration file tampering

If the malware determines that the system it is on is a valid target, its first action is local file tampering. “IncreaseChlorineLevel()” checks a hardcoded list of configuration files associated with desalination, reverse osmosis, chlorine control, and water treatment OT/Industrial Control Systems (ICS).  As soon as it finds any one of these file present, it appends a fixed block of text to it and returns immediately.

The block of text appended to relevant configuration files.
Figure 13: The block of text appended to relevant configuration files.

The appended block of text contains the following entries: “Chlorine_Dose=10”, “Chlorine_Pump=ON”, “Chlorine_Flow=MAX”, “Chlorine_Valve=OPEN”, and “RO_Pressure=80”. Only if none of the hardcoded files are found does the malware proceed to its network-based OT discovery logic.

OT discovery and protocol logic

This section of the code attempts to identify devices on the local subnet, assign each one a protocol label, and then attempt protocol-specific communication. While the overall structure is consistent across protocols, the implementation quality varies significantly.

Figure 14: The ICS scanning function.

The discovery routine, “UZJctUZJctUZJct()”, obtains the local IPv4 address, reduces it to a /24 prefix, and iterates across hosts 1 through 255. For each host, it probes ports 502 (Modbus), 20000 (DNP3), and 102 (S7comm), which the code labels as “Modbus”, “DNP3”, and “S7” respectively if a valid response is received on the relevant port.

The probing is performed in parallel. For every “ip:port” combination, the code creates a task and attempts a TCP connection. The “100 ms” value in the probe routine is a per-connection timeout on “WaitOne(100, ...)”, rather than a delay between hosts or protocols. In practice, this results in a burst of short-lived OT-focused connection attempts across the local subnet.

Protocol validation and device classification

When a connection succeeds, the malware does not stop at the open port. It records the endpoint as an “ICSDevice” with an IP address, port, and protocol label. It then performs a second-stage validation by writing a NULL byte to the remote stream and reading the response that comes back.

For Modbus, the malware checks whether the first byte of the reply is between 1 and 255, for DNP3, it checks whether the first two bytes are “05 64”, and for S7comm, it checks whether the first byte is “03”. These checks are not advanced parsers, but they do show that the author understood the protocols well enough to add lightweight confirmation before sending follow-on data.

 The Modbus read request along with unfinished code for additional protocols.
Figure 15: The Modbus read request along with unfinished code for additional protocols.  

The most developed OT-specific logic is the Modbus-oriented path. In the function “IncreaseChlorineLevel(string targetIP, int targetPort, string parameter)”, the malware connects to the target and sends “01 03 00 00 00 0A”. It then reads the response and parses register values in pairs. The code then uses some basic logic to select a register index: for “Chlorine_Dose”, it looks for values greater than 0 and less than 1000; for “Turbine_Speed”, it looks for values greater than 100.

The Modbus command observed in the sample (01 03 00 00 00 0A) is a Read Holding Registers request. The first byte (0x01) represents the unit identifier, which in traditional Modbus RTU specifies the addressed slave device; in Modbus TCP, however, this value is often ignored or used only for gateway routing because device addressing is handled at the IP/TCP layer.

The second byte (0x03) is the Modbus function code indicating a Read Holding Registers request. The following two bytes (0x00 0x00) specify the starting register address, indicating that the read begins at address zero. The final two bytes (0x00 0A) define the number of registers to read, in this case ten consecutive registers. Taken together, the command requests the contents of the first ten holding registers from the target device and represents a valid, commonly used Modbus operation.

If a plausible register is found, the malware builds a six-byte Modbus write using function code “6” (Write)” and sets the value to 100 for “Chlorine_Dose”, or 0 for any other parameter. If no plausible register is found, it falls back to using hardcoded write frames. In the main malware path, however, the code only calls this function with “Chlorine_Dose".

If none of the ten registers meets the expected criteria, the malware does not abandon the operation. Instead, it defaults to a set of hardcoded Modbus write frames that specify predetermined register addresses and values. This behavior suggests that the attacker had only partial knowledge of the target environment. The initial register-scanning logic appears to be an attempt at dynamic discovery, while the fallback logic ensures that a write operation is still attempted even if that discovery fails.

Incomplete DNP3 and S7comm Logic

The DNP3 and S7comm branches appear much less complete. In “GetCommand()”, the DNP3 path returns the fixed byte sequence “05 64 0A 0C 01 02”, while the S7comm path returns “03 00 00 13 0E 00”. Neither sequence resembles a fully formed command for the respective protocol.

In the case of the S7comm section, the five byte‑ sequence found in the malware sample (05 00 1C 22 1E) most closely matches the beginning of an S7comm parameter block, specifically the header of a “WriteVar (0x05)” request, which is the S7comm equivalent of a Modbus register write operation. In the S7comm protocol, the first byte of a parameter block identifies the function code,  but the remaining bytes in this case do not form a valid item definition. A vaild S7 WriteVar parameter requires at least one item and a full 11-byte variable-specification structure. By comparison this 5‑ byte array is far too short to be a complete or usable command.

The zero item count (0x00) and the trailing three bytes appear to be either uninitialized data or the beginning of an incomplete address field. Together, these details suggest that the attacker likely intended to implement S7 WriteVar functionality, like the Modbus function, but left this portion of the code unfinished.

The DNP3 branch of the malware also appears to be only partially implemented. The byte sequence returned by the DNP3 path (05 64 0A 0C 01 02) begins with the correct two‑byte DNP3 link‑layer sync header (0x05 0x64) and includes additional bytes that resemble the early portion of a link‑layer header. However, the sequence is far too short to constitute a valid DNP3 frame. It lacks the required destination and source address fields, the 16‑bit CRC blocks, and any application‑layer payload in which DNP3 function code would reside. As a result, this fragment does not represent a meaningful DNP3 command.

The incomplete S7 and DNP3 fragments suggest that these protocol branches were still in a developmental or experimental state when the malware was compiled. Both contain protocol‑accurate prefixes, indicating an intent to implement multi‑protocol OT capabilities, however for reasons unknow, these sections were not fully implemented or could not be completed prior to deployment.

USB Propagation

The malware also includes a removable-media propagation mechanism. The “sdfsdfsfsdfsdfqw()” function scans for drives, selects those identified as removable, and copies the hidden payload to each one as “svchost.exe” if it is not already present. The copied executable is marked with the “Hidden” and “System” attributes to reduce visibility.

The malware then calls “CreateUSBShortcut()”, which uses “WScript.Shell” to create .lnk files for each file in the removable drive root. Each shortcut’s TargetPath is set to the hidden malware copy, the icon is set to “shell32.dll, 4” (this is the windows genericfile icon), and the original file is hidden. Were a victim to click this “file,” they would unknowingly run the malware.

Figure 14:The creation of the shortcut on the USB device.

Key Insights

ZionSiphon represents a notable, though incomplete, attempt to build malware capable of malicious interaction with OT systems targeting water treatment and desalination environments.

While many of ZionSiphon’s individual capabilities align with patterns commonly found in commodity malware, the combination of politically motivated messaging, Israel‑specific IP targeting, and an explicit focus on desalination‑related processes distinguishes it from purely opportunistic threats. The inclusion of Modbus sabotage logic, filesystem tampering targeting chlorine and pressure control, and subnet‑wide ICS scanning demonstrates a clear intent to interact directly with industrial processes controllers and to cause significant damage and potential harm, rather than merely disrupt IT endpoints.

At the same time, numerous implementation flaws, most notably the dysfunctional country‑validation logic and the placeholder DNP3 and S7comm components, suggest that analyzed version is either a development build, a prematurely deployed sample, or intentionally defanged for testing purposes. Despite these limitations, the overall structure of the code likely indicates a threat actor experimenting with multi‑protocol OT manipulation, persistence within operational networks, and removable‑media propagation techniques reminiscent of earlier ICS‑targeting campaigns.

Even in its unfinished state, ZionSiphon underscores a growing trend in which threat actors are increasingly experimenting with OT‑oriented malware and applying it to the targeting of critical infrastructure. Continued monitoring, rapid anomaly detection, and cross‑visibility between IT and OT environments remain essential for identifying early‑stage threats like this before they evolve into operationally viable attacks.

Credit to Calum Hall (Cyber Analyst)
Edited by Ryan Traill (Content Manager)

References

1.        https://www.virustotal.com/gui/file/07c3bbe60d47240df7152f72beb98ea373d9600946860bad12f7bc617a5d6f5f/details

Inside the SOC
Darktrace cyber analysts are world-class experts in threat intelligence, threat hunting and incident response, and provide 24/7 SOC support to thousands of Darktrace customers around the globe. Inside the SOC is exclusively authored by these experts, providing analysis of cyber incidents and threat trends, based on real-world experience in the field.
Written by
Calum Hall
Technical Content Researcher

More in this series

No items found.

Blog

/

AI

/

November 19, 2026

Microsoft Perception, Agentic Insiders and the Role of Behavioral Security

Default blog imageDefault blog image

This week at Ignite, Microsoft announced that Microsoft Perception is generally available. It is a significant milestone for the industry, and one we welcome as a Microsoft partner. Microsoft is putting an architectural bet at the center of its security strategy: AI has changed the speed of cyber, and defense must change with it.

At Darktrace, we’ve long held the same view, so it’s encouraging to see a partner of Microsoft’s scale articulate it so clearly and bring it to market.

Security gets a team of agents

Perception is Microsoft’s agentic security system. It orchestrates red, blue, and green team agents through playbooks in a closed loop: red agents simulate attacks, blue agents detect, triage and investigate, and green agents fix and remediate. Rather than adding another assistant to the SOC, it creates coordinated teams of agents that share context. Security teams set the guardrails and approve high-impact actions.

Microsoft frames this as a three-layer cyber stack: signals, sensors and actuators; a context layer; and agents, harness and models on top. Together with partner sources, the foundation produces more than 165 trillion signals per day.  

The announcement also makes a second point that matters just as much. Microsoft is treating AI agents themselves as a new insider-risk category. In its words, “Agents are the new insiders.” They work at machine speed, around the clock, with broad access to data, apps, tools and other agents, and like any insider they can make mistakes, be manipulated or be compromised. New capabilities such as Dynamic Agent Risk (now in preview) and dynamic Conditional Access controls are designed to assess and contain that risk as it changes.

Four truths about security in the agentic era

The announcement is built on four ideas, and together they describe the new reality security teams are defending:

  1. Static security can’t keep up. Microsoft validates what many security practitioners have observed; the window between a vulnerability’s discovery and its exploitation is collapsing from months to minutes. Manual triage can’t answer that pace. Autonomous detection and response is now a requirement, not simply a “nice to have”.
  2. Agents must be treated as insiders. The question for security teams is no longer only “who is this user?” but “what is this agent actually doing?” Identity tells you who an agent is. Policy tells you what it is allowed to do. Neither tells you whether its behavior is normal.
  3. Security is moving from signatures to behavior and context. Dynamic Agent Risk looks across identity, data, endpoint, cloud and network signals and asks whether an agent’s actions are drifting from the intent of its task. That is behavioral thinking, and it is philosophically close to the Adaptive AI approach Darktrace has pioneered.
  4. Machine-speed attacks demand machine-speed response. Humans set the objectives and guardrails. Autonomous systems do the work inside them, both detecting and acting. Microsoft’s design has agents investigate, recommend and, within approved boundaries, act, with security teams setting the limits. At this pace, the gap between knowing something is wrong and doing something about it has to close to seconds.

The opportunity: where control plane meets behavioral context

Perception is an extraordinary control plane for agentic security. The question for security leaders is how to make it even more effective in the environments they actually run.

Real enterprises are heterogeneous. They run AWS and Google Cloud alongside Azure, and OT, IoT and third-party identity alongside Entra. Increasingly they also run AI agents and MCP-connected systems that no single vendor’s telemetry fully covers. Microsoft’s architecture recognizes this, with more than 500 connectors into the wider ecosystem.

That is where Darktrace’s role is distinct. The challenge isn’t signal volume, where Microsoft’s scale is unmatched. It’s deeply understanding what is normal for a specific organization, across every system, identity and agent in it. Together, we deliver two complementary layers:

  • Microsoft brings global intelligence and orchestration. It tells you what is known, what it means and what to do about it.
  • Darktrace brings organization-specific behavioral context. It tells you what is unusual here, right now, across the organizations AI, people and infrastructure, even if it has never been seen before.

The difference shows up in the numbers. A large enterprise generates billions of signals. Darktrace turns them into thousands of incidents, narrows those to hundreds of AI-led investigations, and surfaces as few as 20 that merit a human’s attention. For an agentic SOC, that kind of evaluated, prioritized input makes agents faster, not busier.

As enterprises deploy more agents, the hard problem becomes behavior that is technically permitted but unusual, manipulated or inconsistent with the agent’s purpose. Consider an agent that normally reads Salesforce records during US business hours and suddenly starts enumerating SharePoint and opening unusual outbound connections. Each step may use valid access and approved tools, so identity and policy checks can all pass. Spotting it requires knowing the behavior is abnormal for this agent, in this environment, right now.

That is why one principle in Microsoft’s AI Transparency Initiative matters so much: organizations should be able to see what their AI accesses and does, and independently verify those actions.  

  • Identity and permissions answer “should this agent be allowed to do this?”
  • Behavioral defense answers “even though it is allowed, is what it’s doing normal?”

Deterministic controls around non-deterministic AI are the right foundation, and continuously learned baselines are how organizations can verify those controls are working. ‍

What to look for in an agentic SOC architecture

As organizations adopt platforms like Perception, we’d encourage them to look for three capabilities:

  1. Behavioral context in the control plane. Agents work best with high-confidence signals rather than thousands of raw alerts: here is the behavior that materially deviates from this organization’s learned normal.
  2. ‍Behavioral reasoning on demand. During an investigation, agents should be able to ask what changed around a given identity, device or agent in the last 24 hours, and get a clear answer. ‍
  3. Autonomous, coordinated response across the stack. Response should be as fast as the threat without being reckless: acting immediately when confidence is high, constraining or reversing actions when it isn’t, and keeping a human in the loop when it matters. Within policy and human-approval guardrails, it should target the abnormal behavior rather than the whole business process and span identity, endpoint and the wider environment.

These are the areas where we’re focused, and where we look forward to working with Microsoft and the broader ecosystem.

Response is where speed is won or lost

In the AI era, detection and investigation only matter if action follows quickly. Perception brings agentic orchestration to that problem, with security teams setting the guardrails and approving high-impact steps. Autonomous response from Darktrace takes a complementary approach: it responds on behavioral deviation, in a way that is proportionate and targeted. It can interrupt activity that is abnormal for that specific device, identity or agent while leaving normal business activity running.

The two approaches fit together naturally. Behavioral signals can tell an orchestration layer that something has changed and how confident the system is. Orchestrated playbooks can then coordinate wider containment across identity, endpoint and cloud, within the policies customers define.

The bigger picture

The industry’s architecture is converging on identity, policy, telemetry, behavior, autonomous reasoning and machine-speed response. Microsoft is building formidable pieces of that picture, especially identity, policy, telemetry and orchestration.

Behavior is the layer Darktrace is built to provide: an independent intelligence that continuously learns how humans, machines and AI agents normally act, detects the moment reality departs from intent, and can respond to that deviation in a targeted, proportionate way that keeps normal business running. When attacks move at machine speed, understanding what is abnormal and acting on it quickly have to go together.

Perception brings together what Microsoft knows and orchestrates the response across its stack. Darktrace helps organizations understand what is normal for them and act on what isn’t. Customers building their agentic security strategy will need both.

Congratulations to our partners at Microsoft on this milestone. We look forward to helping customers secure the agentic era together.

Learn more about Darktrace Behavioral Defense.

Continue reading
About the author
Kellie Regan
Director, Product Marketing - Cloud Security

Blog

/

AI

/

November 19, 2026

Agentic attacks in the wild: Inside three ARTEX investigations

Default blog imageDefault blog image

Key insights

·       Darktrace observed activity in several customer environments involving infrastructure and behaviors associated with reported ARTEX-enabled campaigns, including activity linked to China-nexus espionage and financially motivated attacks.

·       In the higher-confidence cases, known infrastructure appeared alongside beaconing, unusual TLS, reconnaissance, tunneling, script retrieval, SMB activity and external data transfer. One Azure case had no observed network command and control at all, only key listing and application identity activity.

·       ARTEX is now used by actors with different objectives. Identifying the framework can help locate relevant activity, while the behavior around the affected system provides the stronger evidence for understanding compromise and attribution.

‍

Introduction

Attackers have moved beyond using AI for phishing and are running agentic frameworks; open-source tools like ARTEX, CAIRN, PentAGI and Hexstrike that let AI agents scan, exploit and move through an environment with little human input [1]. ARTEX has moved quickly from an offensive security project into live intrusion activity. Since at least August 2026, the framework has been associated with espionage, financially motivated attacks and opportunistic campaigns across government, financial services, technology, defense, telecommunications and other sectors. The FBI has documented ARTEX activity spanning reconnaissance, exploitation, credential abuse, lateral movement, collection, command and control, and exfiltration [2].

Darktrace has observed activity in customer environments that overlaps with infrastructure, tooling and behaviors reported in several of these campaigns. The cases below highlight how those intrusions appeared on the network, including the activity visible before an analyst can determine which actor or framework may be responsible.

ARTEX as part of the attacker workflow

ARTEX is an open-source, multi-agent offensive security platform that converts high-level objectives into actions carried out by specialized agents [2]. Operators have used it to automate activity across the attack lifecycle, including scanning, exploitation, credential activity and post-compromise operations. Some reported intrusions moved from initial targeting to privileged access in under three hours [2].

The framework is now associated with actors pursuing different objectives and using different post-exploitation tooling. The original public repository was removed in October 2026, but archived copies and localized versions remain available. For defenders, ARTEX is more useful as context for an investigation than as an attribution shortcut.

Darktrace observations show the same point at network level. Infrastructure linked to reported ARTEX campaigns appeared alongside beaconing, tunneling, unusual TLS, reconnaissance, file retrieval and lateral activity.

Case 1: Espionage-linked activity

Darktrace observed activity associated with BLOODALCHEMY command-and-control (C2) infrastructure in two customer environments. In one environment, five devices communicated with nginxmysql[.]com and 45.153.126[.]223 between October 23 and November 2. Both indicators had previously been linked to China-nexus espionage activity involving ARTEX [4].

The connections were accompanied by suspicious beaconing, repeated successful connections and unusual self-signed TLS. One system began making suspicious TLS connections to nginxmysql[.]com on October 23. Beaconing followed on October 24, and SMB scanning appeared later in the sequence. Other affected systems subsequently communicated with previously unseen external endpoints.

Prior to the observed C2, Darktrace detected curl requests retrieving executable and script content including /systemd, /wazuh, linux_amd64, u.sh and q.sh. Related telemetry included DRSGetNCChanges activity consistent with DCSync, LDAP enumeration, DNS zone transfers, network scanning, certificate downloads and SMB writes involving start.bat.

Figure 1: Autonomous Response model alert in due to the identification of multiple underlying lateral movement related activities, including spikes in SMB and RPC operations on an affected Ivanti network device during an Artex-supported Vapor Panda intrusion.
Figure 2: Autonomous Response action summary console listing all recommended actions for an affected Ivanti network device following the detection of several underlying anomalous behaviors during the Artex-supported Vapor Panda intrusion.

BLOODALCHEMY has been reported in more than one China-nexus cluster [5]. SoftEther infrastructure observed in one environment provided another point of correlation with reported VAPOR PANDA tradecraft. The moderate-confidence assessment tying this to a Chinese nexus threat actor is based on infrastructure, victim context, tradecraft behavior that followed the initial connection, and public reporting.

Case 2: Financial-sector activity

In early November, Darktrace observed activity in two financial-sector environments that overlap with reported ARTEX infrastructure. On November 5, a server connected to 213.111.150[.]145:8099 and downloaded wisp-agent, associated with WISP C2 activity.

In a second environment, a device attempted connections to 85.11.167[.]52 on November 1 and 2. The connections occurred alongside beaconing to new and rare destinations, anomalous TLS, possible tunneling and unusual package downloads.

The device also communicated with temp[.]sh, ssh.localhost[.]run, localhost[.]run, termbin[.]com, pastebin[.]com and GitHub-hosted release infrastructure. The sequence showed potential tooling retrieval, C2, tunneling and use of external services capable of staging or moving data.

Case 3: Credential collection in Azure

On October 20, Darktrace observed a sweep of credential collection in a customer’s Azure environment that is consistent with ARTEX use. The intrusion began with connections to the virtual network from 43.98.196[.]73 (Alibaba Cloud). Open-source scanning data associates the address with ARTEX.

A compromised user identity made requests to the Azure Resource Manager API, though how it was compromised is unknown. The requests listed keys and secrets across API Management, web and function apps, Cognitive Services, Event Hubs, storage accounts and a managed Kubernetes cluster. With those credentials the attacker compromised multiple application identities. The attacker then used the application identities and the user identity to retrieve more credentials, change firewall and security rules and run scripts on virtual machines. Microsoft Defender alerts showed attempted execution of scripts related to Mimikatz and Impacket.

In a cloud environment, such privileged and enumerative operations are potentially indistinguishable from legitimate administrative actions. Consequently, the sequence and context of such operations signified more than any specific API call. There was no C2 traffic to match, and listing keys and using application identities are also what administrators and automation do. A compromised user, credential retrieval across several Azure services, new application identities, more credential collection and changes to security controls only made sense together, and by then access was spreading through the cloud environment.

Figure 3: Darktrace Advanced Search showing Azure AD operation names associated with the activity.
Figure 4: Darktrace Advanced Search showing the Azure Resource Manager action used to execute scripts.
Figure 5: Darktrace Device Event log window showing the triggered models related to the suspicious Azure activity.
Figure 6: Autonomous Response model alert for affected Azure account during Artex intrusion due to the  unusual source IP for the noted activity and the unexpected privileged operations in the Azure environment.
Figure 7: Autonomous Response action summary console for hijacked cloud Azure Identity following the detection of anomalous SaaS/IaaS actions including key/secret requests and Firewall configuration changes.

‍

Shared tooling weakens attribution signals

The three cases looked different on the network: BLOODALCHEMY infrastructure and espionage-consistent tradecraft, WISP infrastructure in financial-sector systems, and Azure identity and control-plane activity with no C2 at all.

Shared tooling has complicated attribution for years, and ARTEX adds to it because the framework carries out more of the attack, so reconnaissance, exploitation attempts and some post-compromise activity can look the same whoever set the objective. Attribution still needs infrastructure, malware, victimology, timing and the sequence of behavior inside the environment, and an ARTEX link is one input to that. Moreover, the behavior and the agent will still be shaped by the biases, preferences, and tradecraft of its operators.

At its core, ARTEX’s planner and worker agents identify and record facts, while exploring intents. However, the way the Large Language Model (LLM) chooses its next steps toward an objective is shaped fundamentally by the skills and tools available in its context. This includes its starting context, as well as various possible Model Context Protocol (MCP) plugins and user-injected context (e.g. user input mid-operation). Operators still have preferred tradecraft, and this preference shapes the pathways the harness will provide to the LLM in context prompts. Even on the same platform, customization based on operator preferences and goals will remain a key facet of campaign and actor identification.

Detection and response

Across the Darktrace cases, the stronger signals came from behavior after the connection: beaconing, abnormal TLS, reconnaissance, unusual authentication, tunneling, script retrieval, SMB activity and outbound transfer. Automated activity also produced compressed bursts of requests, sequential enumeration and repeated attempts with small changes.

With agentic tools reaching privileged access in under three hours, analysts may have only a short window to connect scanning or authentication activity to credential changes, new administrative access, tooling retrieval or lateral movement. Keeping network, identity and cloud telemetry together shows what happened before and after the first match, and whether a connection was noise or an intrusion moving toward privileged access.

In some of these cases, Darktrace's Autonomous Response took or recommended action against the activity, including blocking internal connectivity over specified ports, inhibiting outbound connections with likely C2 endpoints, and/or disabling SaaS/IaaS user accounts. While many of these recommended actions required human confirmation due to deployment configuration preferences, had these actions been applied autonomously at the time of alert, the scope of activity would have been limited, thereby giving analysts time to investigate and respond.

ARTEX won’t be the last of these frameworks. The next ones will come with new capabilities and defaults, but an intrusion still must interact with the target environment, and that shows up in how systems, identities and applications behave even after the indicator has aged out. The harder intrusions to catch will have no infrastructure to match, just ordinary API calls and legitimate identities. Detecting those depends on knowing what normal looks like for each identity and device. Containing abnormal activity at machine speed can stop an intrusion progressing while analysts establish impact and its extent. Autonomous Response does not replace that human analysis, but rather augments defender teams' capabilities by extending the window for investigation.

‍

Credit to Nathaniel Jones (SVP, Global Threat Intelligence), Eugene Chua (Principal Analyst), Sam Lister (Senior Security Researcher), Adam Potter (Associate Principal Cyber Analyst), Daniel Levy (Threat Hunting Data Scientist)

Edited by Ryan Traill (Content Manager)

Appendices

Darktrace model detections

·       Anomalous Connection / Multiple HTTP POSTs to Rare Hostname

·       Anomalous Connection / Rare External SSL Self-Signed

·       Anomalous Connection / Suspicious Self-Signed SSL

·       Compromise / Large Number of Suspicious Successful Connections

·       Compromise / Suspicious Beaconing Behaviour

·       Compromise / Agent Beacon to New Endpoint

·       Compromise / Beacon to Young Endpoint

·       Compromise / SSL Beaconing to Rare Destination

·       Compromise / Suspicious TLS Beaconing To Rare External

·       Compromise / Possible Tunnelling to Bin Services

·       Device / Anomalous Github Download

·       Device / Anomalous Package Download

·       Device / Initial Attack Chain Activity

Indicators of compromise (IoCs)

nginxmysql[.]com

45.153.126[.]223

213.111.150[.]145

wisp-agent

85.11.167[.]52

43.98.196[.]73

192.121.47[.]20

www[.]ddduuuppp[.]com

89.46.234[.]50

sqlserver-sqlite[.]com

199.19.95[.]66

graphics-hook-filter64.dll

graphics-hook_config.bin

References

1.      Google. GTIG AI Threat Tracker - https://cloud.google.com/blog/topics/threat-intelligence/agentic-threat-activity-crosses-a-significant-threshold

2.      Federal Bureau of Investigation, Threat Overview of ARTEX Autonomous Penetration Testing Framework, FLASH-20261113-001, 13 November 2026: https://www.ic3.gov/CSA/2026/261113.pdf

3.      Microsoft Security Blog, Storm-0940 Uses ARTEX Agentic AI in ConnectWise Supply Chain Compromise: https://www.microsoft.com/en-us/security/blog/2026/10/21/storm-0940-uses-artex-agentic-ai-connectwise-supply-chain-compromise/

4.      CrowdStrike, VAPOR PANDA Uses ARTEX to Scale Global Campaign: https://www.crowdstrike.com/en-us/blog/vapor-panda-uses-artex-to-scale-global-campaign/

5.      Bitdefender, SilkParasite: Tracking a China-Nexus APT Across Central Asia: https://www.bitdefender.com/en-us/blog/businessinsights/silkparasite-tracking-china-nexus-apt-across-central-asia

‍

‍

Continue reading
About the author
Nathaniel Jones
SVP, Global Threat Intelligence
Your data. Our AI.
Elevate your network security with Darktrace AI