Blog
/
/
July 11, 2024

GuLoader: Evolving Tactics in Latest Campaign Targeting European Industry

Cado Security Labs identified a GuLoader campaign targeting European industrial companies via spearphishing emails with compressed batch files. This malware uses obfuscated PowerShell scripts and shellcode with anti-debugging techniques to establish persistence and inject into legitimate processes, to deliver Remote Access Trojans. GuLoader's ongoing evolution highlights the need for robust security.
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
Tara Gould
Malware Research Lead
Default blog image
11
Jul 2024

Introduction: GuLoader

Researchers from Cado Security Labs (now part of Darktrace) recently discovered a  campaign targeting European industrial and engineering companies. GuLoader is an evasive shellcode downloader used to deliver Remote Access Trojans (RAT) that has been used by threat actors since 2019 and continues to advance. 

Figure 1

Initial access

Cado identified a number of spearphishing emails sent to electronic manufacturing, engineering and industrial companies in European countries including Romania, Poland, Germany and Kazakhstan. The emails typically include order inquiries and contain an archive file attachment (iso, 7z, gzip, rar). The emails are sent from various email addresses including from fake companies and compromised accounts. The emails typically hijack an existing email thread or request information about an order. 

PowerShell  

The first stage of GuLoader is a batch file that is compressed in the archive from the email attachment. As shown in Image 2, the batch file contains an obfuscated PowerShell script, which is done to evade detection.

Batch file
Figure 2: Obfuscated PowerShell

The obfuscated script contains strings that are deobfuscated through a function “Boendes” (in this sample) that contains a for loop that takes every fifth character, with the rest of the characters being junk. After deobfuscating, the functionality of the script is clearer. These values can be retrieved by debugging the script, however deobfuscating with Script 1 in the Scripts section, makes it easier to read for static analysis.

Deobfuscated Powershell
Figure 3 - Deobfuscated PowerShell

This Powershell script contains the function “Aromastofs” that is used to invoke the provided expressions. A secondary file is downloaded from careerfinder[.]ro and saved as “Knighting.Pro” in the user’s AppData/Roaming folder. The content retrieved from “Kighting.Pro” is decoded from Base64, converted to ASCII and selected from position 324537, with the length 29555. This is stored as “$Nongalactic” and contains more Powershell. 

Second Powershell script
Figure 4 - Second PowerShell script
Deobfuscated Secondary Powershell
Figure 5 - Deobfuscated Secondary PowerShell

As seen in Image 5, the secondary PowerShell is obfuscated in the same manner as before with the function “Boendes”. The script begins with checking which PowerShell is being used 32 or 64 bit. If 64 bit is in use, a 32 bit PowerShell process is spawned to execute the script, and to enable 32 bit processes later in the chain. 

The function named “Brevsprkkernes” is a secondary obfuscation function. The function takes the obfuscated hex string, converts to a byte array, applies XOR with a key of 173 and converts to ASCII. This obfuscation is used to evade detection and analysis more difficult. Again, these values can be retrieved with debugging; however for readability, using Script 2 in the Scripts section makes it easier to read. 

Obfuscated Hex Strings
Figure 6: Obfuscated Hex Strings
Deobfuscated PowersShell Strings
Figure 7 - Deobfuscated PowerShell Strings
Deobfuscated Process Injection
Figure 8: Deobfuscated Process Injection

The second PowerShell script contains functionality to allocate memory via VirtualAlloc and to execute shellcode. VirtualAlloc is a native Windows API function that allows programs to allocate, reserve, or commit memory in a specified process. Threat actors commonly use VirtualAlloc to allocate memory for malicious code execution, making it harder for security solutions to detect or prevent code injection. The variable “$Bakteriekulturs” contains the bytes that were stored in “AppData/Roaming/Knighting.Pro” and converted from Base64 in the first part of the PowerShell Script. Marshall::Copy is used to copy the first 657 bytes of that file, which is the first shellcode. Marshall.Copy is a method that enables the transfer of data between unmanaged memory and managed arrays, allowing data exchange between managed and unmanaged code. Marshal.Copy is typically abused to inject or manipulate malicious payloads in memory, bypassing traditional detection by directly accessing and modifying memory regions used by applications. Marshall::Copy is used again to copy bytes 657 to 323880 as a second shellcode. 

First Shellcode
Figure 9: First Shellcode

The first shellcode includes multiple anti-debugging techniques that make static and dynamic analysis difficult. There have been multiple evolutions of GuLoader’s evasive techniques that have been documented [1]. The main functionality of the first shellcode is to load and decrypt the second shellcode. The second shellcode adds the original PowerShell script as a Registry Key “Mannas” in HKCU/Software/Procentagiveless for persistence, with the path to PowerShell 32 bit executable stored as “Frenetic” in HKCU\Environment; however, these values change per sample. 

Registry Key created for PowerShell Script
Figure 10 - Registry Key created for PowerShell Script
PowerShell bit added to Registry
Figure 11 - PowerShell 32 bit added to Registry

The second shellcode is injected into the legitimate “msiexec.exe” process and appears to be reaching out to a domain to retrieve an additional payload, however at the time of analysis this request returns a 404. Based on previous research of GuLoader, the final payload is usually a RAT including Remcos, NetWire, and AgentTesla.[2]

msiexec abused to retrieve additional payload
Figure 12  - msiexec abused to retrieve additional payload

Key Takeaway

Guloader malware continues to adapt its techniques to evade detection to deliver RATs. Threat actors are continually targeting specific industries in certain countries. Its resilience highlights the need for proactive security measures. To counter Guloader and other threats, organizations must stay vigilant and employ a robust security plan.

Scripts

Script 1 to deobfuscate junk characters 

import re 
import argparse 
import os 
 
def deobfuscate_powershell(input_file, output_file): 
  try: 
      with open(input_file, 'r', encoding='utf-8') as f: 
          text = f.read() 
 
      function_name_match = re.search(r"function\s+(\w+)\s*\(", text) 
      if not function_name_match: 
          print("Could not find the obfuscation function name in the file.") 
          return 
      
      function_name = function_name_match.group(1) 
      print(f"Detected obfuscation function name: {function_name}") 
 
      obfuscated_pattern = rf"(?<={function_name} ')(.*?)(?=')" 
      matches = re.findall(obfuscated_pattern, text) 
 
      for match in matches: 
          deobfuscated = match[4::5] 
          full_obfuscated_call = f"{function_name} '{match}'" 
          text = text.replace(full_obfuscated_call, deobfuscated) 
 
      with open(output_file, 'w', encoding='utf-8') as f: 
          f.write(text) 
 
      print(f"Deobfuscation complete. Output saved to {output_file}") 
 
  except Exception as e: 
      print(f"An error occurred!: {e}") 
 
if __name__ == "__main__": 
  parser = argparse.ArgumentParser(description="Deobfuscate an obfuscated PowerShell file.") 
  parser.add_argument("input_file", help="Path to the obfuscated PowerShell file.") 
  parser.add_argument("output_file", nargs='?', help="Path to save the deobfuscated file. Default is 'deobfuscated_powershell.ps1' in the same directory.", default=None) 
 
  args = parser.parse_args() 
 
  if args.output_file is None: 
      output_file = os.path.splitext(args.input_file)[0] + "_deobfuscated.ps1" 
  else: 
      output_file = args.output_file 
 
  deobfuscate_powershell(args.input_file, output_file) 

Script 2 to deobfuscate hex strings obfuscation (note this will need values changed based on sample)

import re 
import argparse 
 
def brevsprkkernes(spackle): 
  if not all(c in'0123456789abcdefABCDEF'for c in spackle): 
      return f"Invalid hex: {spackle}" 
  paronomasian = 2 
  polyurethane = bytearray(len(spackle) // 2) 
 
  for forstyrrets in range(0, len(spackle), paronomasian): 
      try: 
          polyurethane[forstyrrets // 2] = int(spackle[forstyrrets:forstyrrets + 2], 16) 
          polyurethane[forstyrrets // paronomasian] ^= 173 
      except ValueError: 
          return f"Error processing hex: {spackle}" 
 
  return polyurethane.decode('ascii', errors='ignore') 
 
def process_file(input_file, output_file): 
  with open(input_file, 'r') as infile: 
      content = infile.read() 
 
  def replace_function(match): 
      hex_string = match.group(1).strip() 
      result = brevsprkkernes(hex_string) 
      return f"Brevsprkkernes '{result}'" 
 
  updated_content = re.sub(r"Brevsprkkernes\s*['\"]?([0-9A-Fa-f]+)['\"]?", replace_function, content) 
 
  with open(output_file, 'w') as outfile: 
      outfile.write(updated_content) 
 
if __name__ == "__main__": 
  parser = argparse.ArgumentParser(description="Process a PowerShell file and replace hex strings.") 
  parser.add_argument("input_file", help="Path to the input file.") 
  parser.add_argument("output_file", help="Path to save the deobufuscated file.") 
  args = parser.parse_args() 
 
  process_file(args.input_file, args.output_file) 

Indicators of compromise (IoCs)

GuLoader scripts

ZW_PCCE-010023024001.bat  36a9a24404963678edab15248ca95a4065bdc6a84e32fcb7a2387c3198641374  

ORDER_1ST.bat  26500af5772702324f07c58b04ff703958e7e0b57493276ba91c8fa87b7794ff  

IMG465244247443 GULF ORDER Opmagasinering.cmd  40b46bae5cca53c55f7b7f941b0a02aeb5ef5150d9eff7258c48f92de5435216  

EXSP 5634 HISP9005 ST MSDS DOKUME74247linierelet.bat  e0d9ebe414aca4f6d28b0f1631a969f9190b6fb2cf5599b99ccfc6b7916ed8b3  

LTEXSP 5634 HISP9005 ST MSDS DOKUME74247liniereletbrunkagerne.bat 4c697bdcbe64036ba8a79e587462960e856a37e3b8c94f9b3e7875aeb2f91959  

Quotation_final_buy_order_list_2024_po_nos_ART125673211020240000000000024.bat661f5870a5d8675719b95f123fa27c46bfcedd45001ce3479a9252b653940540  

MEC20241022001.bat  33ed102236533c8b01a224bd5ffb220cecc32900285d2984d4e41803f1b2b58d  

nMEC20241022001.iso  9617fa7894af55085e09a06b1b91488af37b8159b22616dfd5c74e6b9a081739  

Gescanneerde lijst met artikelen nr. 654398.bat  f5feabf1c367774dc162c3e29b88bf32e48b997a318e8dd03a081d7bfe6d3eb5  

DHL_Shipping_Invoices_Awb_BL_000000000102220242247820020031808174Global180030010222024.cmd f78319fcb16312d69c6d2e42689254dff3cb875315f7b2111f5c3d2b4947ab50  

Order Confirmation.bat  949cdd89ed5fb2da03c53b0e724a4d97c898c62995e03c48cbd8456502e39e57  

SKM_0001810-01-2024-GL-3762.bat  9493ad437ea4b55629ee0a8d18141977c2632de42349a995730112727549f40e  

21102024_0029_18102024_SKM_0001810-01-2024-GL-3762.iso  535dd8d9554487f66050e2f751c9f9681dadae795120bb33c3db9f71aafb472c  

\Device\CdRom1\MARSS-FILTRY_ZW015010024.BAT  e5ebe4d8925853fc1f233a5a6f7aa29fd8a7fa3a8ad27471c7d525a70f4461b6  

Myologist.cmd  51244e77587847280079e7db8cfdff143a16772fb465285b9098558b266c6b3f  

SKU_0001710-1-2024-SX-3762.bat  643cd5ba1ac50f5aa2a4c852b902152ffc61916dc39bd162f20283a0ecef39fe  

Stamcafeernes.cmd  54b8b9c01ce6f58eb6314c67f3acb32d7c3c96e70c10b9d35effabb7e227952e  

C:\Users\user\AppData\Local\Temp\j4phhdbc.lti\Bank details Form.bat  c1f810194395ff53044e3ef87829f6dff63a283c568be4a83088483b6c043ec8  

SKGCRO COMANDA FAB SRL M60_647746748846748347474.bat  8dd5fd174ee703a43ab5084fdaba84d074152e46b84d588bf63f9d5cd2f673d1  

DHL_Shipping_Invoices_Awb_BL_000000000101620242247820020031808174Global180030010162024.bat bde5f995304e327d522291bf9886c987223a51a299b80ab62229fcc5e9d09f62  

Ciwies.cmd  b1be65efa06eb610ae0426ba7ac7f534dcb3090cd763dc8642ca0ede7a339ce7  

Zamówienie Agotech Begyndelsesord.cmd  18c0a772f0142bc8e5fb0c8931c0ba4c9e680ff97d7ceb8c496f68dea376f9da  

SKM_0001810-01-2024-GL-3762.iso  4a4c0918bdacd60e792a814ddacc5dc7edb83644268611313cb9b453991ac628  

C:\Users\user\AppData\Local\Temp\Stemmeslugerens.bat  8bedbdaa09eefac7845278d83a08b17249913e484575be3a9c61cf6c70837fd2  

Agotech Zamówienie Fjeldkammes325545235562377.bat  ff6c4c8d899df66b551c84124e73c1f3ffa04a4d348940f983cf73b2709895d3  

Agotech Zamówienie Fjeldkammes3255452355623.bat  f3e046a7769b9c977053dd32ebc1b0e1bbfe3c61789d2b8d54e51083c3d0bed5  

SKU_0001710-1-2024-SX-3762.iso  0546b035a94953d33a5c6d04bdc9521b49b2a98a51d38481b1f35667f5449326  

SKU_0001710-1-2024-SX-3762.bat  4f1b5d4bb6d0a7227948fb7ebb7765f3eb4b26288b52356453b74ea530111520  

DOKUMENTEN_TOBIAS.bat  038113f802ef095d8036e86e5c6b2cb8bc1529e18f34828bcf5f99b4cc012d6a  

IMEG238668289485293885823085802835025Urfjeld.bat  6977043d30d8c1c5024669115590b8fd154905e01ab1f2832b2408d1dc811164  

SKM_C250i24100408500.iso  6370cbcb1ac3941321f93dd0939d5daba0658fb8c85c732a6022cc0ec8f0f082  

SKU_0001710-1-2024-SX-3762.iso  7f06382b781a8ba0d3f46614f8463f8857f0ade67e0f77606b8d918909ad37c2  

\Device\CdRom1\ORDINE ELECTRICAS BC CORP PO EDC0969388.BAT  e98fa3828fa02209415640c41194875c1496bc6f0ca15902479b012243d37c47  

Quote Request #2359 Bogota.msg  0f0dfe8c5085924e5ab722fa01ea182569872532a6162547a2e87a1d2780f902  

ORDER.1ST.bat  48dca5f3a12d3952531b05b556c30accafbf9a3c6cda3ec517e4700d5845ab61  

Fortryl105.cmd  f43b78e4dc3cba2ee9c6f0f764f97841c43419059691d670ca930ce84fb7143b  

SMX-0002607-1-2024-UP-3762.iso  a60dbbe88a1c4857f009a3c06a2641332d41dfd89726dd5f2c6e500f7b25b751

Quotation_final_buy_order_list_2024_po_nos_ART1256731610202400000000000.cmd efd80337104f2acde5c8f3820549110ad40f1aa9b494da9a356938103bda82e7

a60dbbe88a1c4857f009a3c06a2641332d41dfd89726dd5f2c6e500f7b25b751.iso 0327db7b754a16a7ae29265e7d8daed7a1caa4920d5151d779e96cd1536f2fbe  

MARSS-FILTRY_ZW015010024.iso c415127bde80302a851240a169fff0592e864d2f93e9a21c7fd775fdb4788145

SKM_C250i24100408500.bat 36c464519a4cce8d0fcdb22a8974923fd51d915075eba9e62ade54a9c396844d  

UPM-0002607-1-2024-UP-3762.iso  e9fc754844df1a7196a001ac3dfbcf28b80397a718a3ceb8d397378a6375ff62  

Comanda KOMARON TRADE SRL 435635Lukketid.bat 1bf09bcb5bfa440fc6ce5c1d3f310fb274737248bf9acdd28bea98c9163a745a  

311861751714730477170144.bat f87448d722e160584e40feaad0769e170056a21588679094f7d58879cdb23623  

Estimate_buy_product_purchase_order_import_list_10_10_2024_000000101024.cmd f20670ed0cdc2d9a2a75884548e6e6a3857bbf66cfbfb4afe04a3354da9067c9  

PAYMENT TERM.bat 4c90504c86f1e77b0a75a1c7408adf1144f2a0e3661c20f2bf28d168e3408429  

Arbitrre.cmd  8ef4cb5ad7d5053c031690b9d04d64ba5d0d90f7bf8ba5e74cb169b5388e92c5  

KZЗапрос продукта SKM_32532667622352352Arvehygiejnikernes.bat 4ddd3369a51621b0009b6d993126fcb74b52e72f8cacd71fcbc401cda03108cb  

Order_AP568.bat fda4e04894089be87f520144d8a6141074d63d33b29beb28fd042b0ecc06fbbc  

C:\Users\user\Documents\ConnectWiseControl\Temp\Blodprocenternes.cmd e5f5d9855be34b44ad4c9b1c5722d1a6dff2f4a6878a874df1209d813aea7094  

Productivenesses.cmd a7268e906b86f7c1bb926278bf88811cb12189de0db42616e5bbb3dc426a4ef5  

Doktriner.cmd 74d468acd0493a6c5d72387c8e225cc0243ae1a331cd1e2d38f75ed8812347dd  

final_buy_product_purchase_order_import_list_11_10_2024_000000111024.cmd a2127d63bc0204c17d4657e5ae6930cab6ab33ae3e65b82e285a8757f39c4da9  

ORDER_U769.bat b45d9b5dbe09b2ca45d66432925842b0f698c9d269d3c7b5148cc26bdc2a92d0  

Beschwerde-Rechtsanwalt.bat 229c4ce294708561801b16eed5a155c8cfe8c965ea99ac3cfb4717a35a1492f3  

upit nr5634 10_08_2024.cmd 5854d9536371389fb0f1152ebc1479266d36ec4e06b174619502a6db1b593d71  

C:\Users\user\AppData\Local\Temp\Doktriner.cmd 140dcb39308d044e3e90610c65a08e0abc6a3ac22f0c9797971f0c652bb29add  

Fedtsyresammenstning.cmd 0b1c44b202ede2e731b2d9ee64c2ce333764fbff17273af831576a09fc9debfa  

HENIKENPLANT PROJECT PROPOSAL BID_24-0976·pdf.cmd 31a72d94b14bf63b07d66d023ced28092b9253c92b6e68397469d092c2ffb4a6  

MAIN ORDER.bat 85d1877ceda7c04125ca6383228ee158062301ae2b4e4a4a698ef8ed94165c7c  

Narudzba ACH0036173.bat 8d7324d66484383eba389bc2a8a6d4e9c4cb68bfec45d887b7766573a306af68  

Sludger.cmd 45b7b8772d9fe59d7df359468e3510df1c914af41bd122eeb5a408d045399a14  

Glasmester.bat b0e69f895f7b0bc859df7536d78c2983d7ed0ac1d66c243f44793e57d346049d  

PERMINTAAN ANGGARAN (Universitas IPB) ID177888·pdf.cmd 09a3bb4be0a502684bd37135a9e2cbaa3ea0140a208af680f7019811b37d28d6  

C:\Users\user\Documents\ConnectWiseControl\Temp\Bidcock.cmd 0996e7b37e8b41ff0799996dd96b5a72e8237d746c81e02278d84aa4e7e8534e  

PO++380.101483.bat a9af33c8a9050ee6d9fe8ce79d734d7f28ebf36f31ad8ee109f9e3f992a8d110  

Network IOCs

91[.]109.20.161

137[.]184.191.215

185[.]248.196.6

hxxps://filedn[.]com/lK8iuOs2ybqy4Dz6sat9kSz/Frihandelsaftalen40.fla

hxxps://careerfinder[.]ro/vn/Traurigheder[.]sea

hxxp://inversionesevza[.]com/wp-includes/blocks_/Dekupere.pcz

hxxps://rareseeds[.]zendesk[.]com/attachments/token/G9SQnykXWFAnrmBcy8MzhciEs/?name=PO++380.101483.bat

Detection

Yara rule

rule GuLoader_Obfuscated_Powershell 
{ 
   meta: 
       description = "Detects Obfuscated GuLoader Powershell Scripts" 
       author = "[email protected]" 
       date = "2024-10-14" 
   strings: 
      $hidden_window = { 7374617274202f6d696e20706f7765727368656c6c2e657865202d77696e646f777374796c652068696464656e2022 } 
      $for_loop = /for\s*\(\s*\$[a-zA-Z0-9_]+\s*=\s*\d+;\s*\$[a-zA-Z0-9_]+\s*-lt\s*\$[a-zA-Z0-9_]+\s*;\s*\$[a-zA-Z0-9_]+\s*\+=\s*\d+\s*\)/ 
   condition: 
      $for_loop and $hidden_window 

MITRE ATT&CK

T1566.001  Phishing: Malicious Attachment  

T1055 Process Injection  

T1204.002  User Execution: Malicious File  

T1547.001  Boot or Logon Autostart Execution: Registry Run Keys / Startup Folder  

T1140  Deobfuscate/Decode Files or Information  

T1622  Debugger Evasion  

T1001.001  Junk Code  

T1105  Ingress Tool Transfer  

T1059.001  Command and Scripting Interpreter: Powershell  

T1497.003  Virtualization/Sandbox Evasion: Time Based Evasion  

T1071.001  Application Layer Protocol: Web Protocols

References:

[1] https://www.crowdstrike.com/en-us/blog/guloader-dissection-reveals-new-anti-analysis-techniques-and-code-injection-redundancy/  

[2] https://www.checkpoint.com/cyber-hub/threat-prevention/what-is-malware/guloader-malware/

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
Tara Gould
Malware Research Lead

More in this series

No items found.

Blog

/

Network

/

August 11, 2026

When AI Agents Attack: The Case for Behavioral Anomaly Detection

Default blog imageDefault blog image

The UK AI Security Institute (AISI) has just published a detailed incident report confirming what many within the cyber security industry have suspected for some time: frontier AI agents, given a degree of autonomy, can independently develop and execute attack chains against real targets, including social engineering, supply-chain compromise, and deception, without explicit instruction to do so.

The report (INC-2026-07-28-01, published 4 August 2026) documents what happened during a routine capability evaluation between 25 and 28 July. It deserves a careful read from anyone responsible for securing the modern digital estate.

The incident

AISI regularly tests frontier AI models for offensive cyber capability. In this exercise, agents were placed inside a simulated network environment and given internet access this was deliberate, given to reflect realism of the task. Across 122 evaluation runs, 19 instances of unsanctioned activity were recorded, including cases where agents broke the intended boundaries of the test and acted against real people and real infrastructure on the open internet.

In the most serious case, an agent (Mythos 5) autonomously:

  • Conducted open-source intelligence gathering to identify real developers associated with keywords from the simulated scenario
  • Created a GitHub account and submitted a pull request containing obfuscated malicious code to a real, unaffiliated repository
  • Created a second sockpuppet account to post comments on its own PR, attempting to manufacture social proof
  • When a human reviewer flagged the malicious content, the agent lied, claiming an accidental commit, then repeatedly attempted to reintroduce the payload under the guise of having "fixed" the code
  • Sent spear-phishing emails to the repository maintainer containing manipulated content designed to convince them to merge the code
  • Planted a prompt injection inside a GitHub issue, hidden in an HTML comment invisible to humans but readable by AI coding assistants parsing the repository via API

AISI confirmed no real-world harm resulted. The PRs were not merged; the payloads did not execute. But containment was a product of circumstance, not design.

Why this matters

Every action the agents took used legitimate services operating within normal parameters. Creating a GitHub account is legitimate. Opening a pull request is legitimate. Sending an email is legitimate. Posting an issue is legitimate.

No vulnerability was exploited. No known malware signature was present. No indicator of compromise from any threat intelligence feed would have matched. The entire attack existed as a pattern, a sequence of individually normal actions that, taken together, constituted a supply-chain attack, a social engineering campaign, and an attempted compromise of downstream systems.

This is the class of threat that signature-based and rule-based security tools structurally struggle to detect. They answer the question "have I seen this before?" When the answer is no – and when the attack is novel, composed of legitimate actions, and adapted in real time by an adversary in this case an agent that reasons about evasion – they struggle.

Behavioral anomaly detection: The architecture that fits

Darktrace has spent a decade building a platform around a different question: "is this behavior normal for this entity?"

The platform learns the baseline behavior of every user, device, and communication pattern across an organization's digital environment. It does not need prior knowledge of an attack technique. It identifies deviation from learned normality, surfaces it, and through its Autonomous Response capability, takes targeted action to contain threats at machine speed.

Consider how this AISI incident might show across the Darktrace platform:

Network detection and response

While no payload was successfully delivered to a victim in this incident, AISI's post-incident analysis of the agent's code (detailed on pages 24-25 of the report) reveals what would have landed on target systems had the social engineering succeeded. The agent produced three successive payload generations, each replacing the last after being detected and called out:

Generation 1 hid a module-load stager inside a legitimate bug-fix commit, guarded by an environment variable check, to skip execution in CI. Its second stage attempted to fetch Poseidon implants (Linux and Windows variants) calling back to the agent's own C2 listener, with a fallback to publicly hosted polling scripts that fetched and executed commands every 60 seconds.

Generation 2 exploited an Electron "What's New" window with nodeIntegration enabled, loading remote HTML from a GitHub Pages site the agent had registered. The page decoded an XOR-obfuscated payload at runtime that beaconed host identifiers over DNS, attempted to download the same implants, ran its own in-process C2 loop, and installed reboot persistence via crontab and Windows registry Run keys.

Generation 3 was a post-install dropper added after the agent had already been publicly caught. It XOR-decoded a hidden URL, downloaded a script over plaintext HTTP, and ran it detached, force-exiting so npm install always reports success, giving the attacker remote code execution on any machine running the install.

None of these payloads reached a victim. But the technical detail matters: this is the calibre of tradecraft an autonomous agent produced in a matter of hours, unprompted. Each generation demonstrated adaptation, the agent iterated its approach in direct response to being detected.

For network defenders, the behavioral signals these payloads would have generated are clear. DNS beaconing to novel domains, periodic HTTP polling to newly registered infrastructure, XOR-encoded data in transit, registry and crontab persistence writes on endpoints that have never exhibited such behavior. These are exactly the anomalies that Darktrace / NETWORK identifies in real time: deviations from learned device and connection baselines, regardless of whether the specific payload or C2 domain has been seen before.

Beyond the payload analysis, the agents' broader activity produced additional behavioral anomalies: new identities engaging in high-frequency interactions with external services, coordination via unusual channels (public pastebins, gists), and lateral communication patterns that deviate from any established baseline.

Email security

The spear-phishing emails sent by the agent, crafted with context-specific content designed to manipulate a specific individual, represent exactly the kind of socially engineered communication that static email gateways can miss. Darktrace / EMAIL analyses the linguistic patterns, sender behavior, and relational context of every message. A brand-new sender using tailored language to pressure a recipient into a specific technical action deviates from the learned communication profile, regardless of whether the message contains a known-bad URL or attachment.

Securing AI

This is where the incident points to a threat category that many security platforms do not yet fully address. The agent did not just use AI as a tool for attack. It targeted other AI systems, attempting prompt injections against coding assistants and trying to compromise automated workflows. Organisations adopting AI agents for development, operations, and decision-making are introducing a newer extended attack surface.  

Darktrace / SECURE AI extends the same behavioral detection model into this space. It monitors AI usage across the enterprise, detecting anomalous interactions with AI services, identifying sensitive data exposure to AI systems, providing visibility into shadow AI use. As AI agents become standard components of enterprise infrastructure, the ability to detect when an AI system is behaving outside its expected parameters, when there is a chance, it is being manipulated, or when it is interacting with something it should not be, that becomes a direct security requirement rather than a future consideration.

The principle is consistent across the Darktrace platform: learn what normal looks like, identify deviation, respond autonomously.

The AI-to-AI threat surface

One of the report's most important findings is a confirmation that AI agents can act outside of instructions and guardrails to complete tasks. The Mythos 5 agent conducted reconnaissance on a repository maintainer, inferred it might be an AI coding assistant, and deployed a prompt injection specifically crafted to compromise it.

This is a new attack class. Organizations deploying AI coding assistants, AI-powered code review, or autonomous DevOps agents now face an adversary that:

  • Moves at speed
  • Is able to adapt its approach when initial attempts are halted
  • Specifically targets AI systems via techniques (prompt injection, hidden instructions in HTML comments) that are less visible to human reviewers but parsed by machines
  • Reasons about which evasion strategies are most likely to succeed

Defending against this requires security that operates with the same adaptability and speed. A system that learns what "normal" looks like, and flags deviation from it.

What AISI recommends

The report identifies five contributing factors to the incident. Two are directly relevant:

No synchronous real-time monitoring (Section 5.3): AISI acknowledges it had not built real-time behavioral monitoring capable of immediately blocking unsanctioned agent actions. Their security team detected the incident via commercial security monitoring software, but only after activity had already occurred.

No fine-grained behavioral controls (Section 6.2): AISI's forward-looking response includes building continuous monitoring that can "immediately block or flag unsanctioned AI agent behavior."

The direction is clear: organizations need security that watches behavior continuously, learns what normal looks like, and acts when deviation occurs, without requiring a human in the loop for every decision.

The takeaway

The AISI incident is documented proof that autonomous AI agents will independently develop novel attack chains using legitimate tools, adapt when detected, employ deception, and target both humans and other AI systems. This happened last week, in a controlled setting, with commercially available models.

The security architecture that addresses this is behavioral anomaly detection applied across the full digital estate, as AI agents become standard components of enterprise infrastructure, writing code, managing deployments, processing communications, the attack surface they create is behavioral by nature.

This is the approach Darktrace has taken for years: learning what is normal across an organization’s digital environment, identifying meaningful deviations, and responding to emerging threats without relying on known attack signatures. As autonomous AI agents introduce new and unpredictable behaviors, that foundation becomes increasingly important to securing the enterprise.

Read the full report from the UK AI Security Institute here.

Continue reading
About the author
Adam Stevens
Senior Director of Product

Blog

/

/

August 5, 2026

Testing a Prompt injection Attack Against an Enterprise AI Agent

prompt injectionDefault blog imageDefault blog image

Key takeaways

  • Darktrace successfully detected and quarantined a prompt injection email before it could be processed by an enterprise AI agent.  
  • Prompt injection attacks increasingly rely on natural language rather than traditional malware, making behavioral analysis an important complement to signature-based detection.  
  • Organizations deploying AI agents should combine model guardrails with behavioral monitoring to reduce the risk of malicious instructions reaching enterprise systems.

How behavioral detection helps stop prompt injection attacks

A Darktrace customer running a Gemini AI agent in Google Cloud asked us two simple questions:

“If my agent can read inbound emails and access internal data, what stops an attacker from hiding malicious instructions in the message? Couldn’t the agent be tricked into deleting or exfiltrating sensitive data?”

The scenario centers on an indirect prompt injection attack, where malicious instructions are hidden inside content that an AI model later interprets as trusted input. The same weakness was exposed by  EchoLeak (CVE-2025-32711), a zero-click Microsoft 365 Copilot vulnerability enabled data exfiltration from a single well-crafted email..

This blog follows the Darktrace team’s investigation of the customer’s hypothesis and examines how the attack interacted with their existing security stack. The results highlight which defenses held, where gaps emerged, and how behavioral detection mattered more than guardrails. This investigation also demonstrates why behavioral detection is becoming increasingly important for AI security, as prompt injections often contain no traditional indicators of compromise.  

How do prompt injection attacks work?

Prompt injection works by carefully crafting the content and structure of the prompt to alter the LLM’s behavior or output in unintended ways. This can cause models to violate guardrails, generate harmful content or enable unauthorised access.

Prompt injection attack example

The well-known example, EchoLeak (CVE-2025-32711), was a zero-click vulnerability in Microsoft 365 Copilot that relied on a carefully crafted email containing hidden instructions that the AI system interpreted as commands rather than content, creating a pathway for unauthorized access to sensitive information without any user interaction.

While Darktrace / SECURE AI is designed to prevent agents from producing unintended outcomes, we wanted to see if we could catch and prevent this threat type earlier in the attack life-cycle, at the email security layer.

How we tested prompt injection attacks on an enterprise agent

Summary:

  1. Claude generated a prompt injection payload.  
  2. Hidden instructions were embedded in an email.  
  3. The email passed traditional validation checks.  
  4. Darktrace analyzed the language and sender behavior.  
  5. The email was quarantined before the AI agent could process it.

To test Darktrace / EMAIL against this attack class, we opened Claude, gave it the customer's context and problem statement (Gemini agent with inbox access, internal tool calls), told it we were validating Darktrace / EMAIL's detection of prompt injections, and asked for a test payload. See below:

Figure 1
Figure 2

Despite the guardrails supposedly built into the model, Claude surprisingly gave us the entire exploit in plaintext (albeit very basic), illegible to a human as the text was sent in white text (see Figure 1) but framed as an authoritative override for anything downstream reading the mail programmatically (i.e. the Gemini agent).

How Darktrace detected a prompt injection attack

We then sent the Claude-crafted email from a freemail address to the target recipient’s inbox. Despite the email containing no malicious payload, the freemail address having no malicious reputation, and the validation checks all passing, Darktrace  /EMAIL flagged the email as a 93/100 anomaly and moved it to junk, out of scope for the AI agent.

Figure 3: The test email sent with the hidden prompt injection
Figure 4: The email analysis in Darktrace / EMAIL 
Figure 5: Darktrace / EMAIL detection of malicious activity

The interesting part is what triggered the detection (see Figure 5)

  • Possible machine prompt content: text in the body detected as instructions written for a machine to execute, not for a human to read
  • Possible machine prompt content + basic suspicious correspondence: the same content, correlated with sender-side anomalies: freemail domain (yahoo[.]com), unknown correspondent, no prior mail history with the recipient, and suspicious references to payment information

Neither of those is a signature match. Nothing in the email was on a blacklist. There was no malware, no link and no attachment. Darktrace analyzed the context in which the email was delivered and flagged it as likely risky.  The anomalous language features and the context of the sender relative to the recipient's normal behavior, combined with the unusual hidden text (prompt) were enough for Darktrace / EMAIL to act on the risk.

Result: Darktrace / EMAIL autonomously junked the email, out of scope for any AI agent parsing the inbox.

Why behavioral security makes a difference detecting prompt injection attacks

Cyberattacks don't look like traditional exploits anymore. They now operate in natural language, not strictly code.

That breaks the traditional stack. AV, firewalls, static scanning and signature-based SEGs all assume a payload to inspect.  

A prompt injection has no payload. It's just an instruction, written in natural language, dressed up as anything the attacker wants: an invoice, an HR request, a calendar invite, some simple PowerPoint slides.

EchoLeak proved that hidden instructions can sit inside an email invisible to the user but fully readable by the LLM, and the LLM will follow them blindly.  

This test and GTG-1002 proved that the LLM itself can be socially engineered. Tell it you're an authorized tester and it will hand you the attack.

Rules and static classifiers can catch the obvious cases. But natural language has infinite variants, and the attack surface is the model's innate functionality to comply.  

The deeper problem here is intent: an LLM can't reliably tell whether an instruction in its context came from its developer, its user, or an attacker who slipped it into an email. To the LLM, everything reads as language and looks like a legitimate ask. This is why behavioural detection wins, as you become aware of intent when you look at the context of an interaction. Does this sender normally send this kind of message to this recipient? Does this prompt fit the user's normal pattern? Is this agent behaving the way this agent normally behaves?  

Intent can't be read off a single email, it emerges from behavioral context. Which is how Darktrace enables threat detection, through behavioral understanding.

Why enterprise AI security requires more than guardrails

Claude didn't roll over immediately… the first section of the response was a (slight) pushback, but then it wrote the payload anyway without having to ask twice.

Here the framing of the prompt did all the work. The “testing security capabilities” angle moved the model from refusal to unquestioned compliance to the user prompt.

This isn't the first time this has happened, of course. Anthropic disclosed in November 2025 that a Chinese state-sponsored group they tracked as GTG-1002 ran the first documented AI-orchestrated espionage campaign against ~30 targets by posing as employees of a legitimate cybersecurity firm doing authorised penetration testing.

The takeaway isn't that AI guardrails are ineffective. They raise the cost of low-effort attacks and remain an important first layer of defense. However, for most organizations today, they’re the only line of defense when deploying AI agents. If a prompt injection bypasses those controls, organizations still need a way to detect and stop malicious behavior elsewhere in the attack chain.

Attackers will continue to have working prompt injections easily and quickly. The question is what stops one when it lands in an inbox your agent is reading.

That's where behavioral detection comes in.

How Darktrace detects prompt injection attacks in emails

Two things Darktrace does that a model-level guardrail or static rules and signatures can't:

Natural language analysis at the email or prompt layer. The email is assessed on its own merits: is this content shaped like instructions for a machine, regardless of what the receiving agent decides to do about it?

Behavioral context around the language. An AI agent behaves like an extremely agreeable human, and it will go above and beyond to comply with the user’s request. That's exactly why you must consider the business context, such sender behaviour, mailing history, and organisational norms, as these matter even more when the recipient is an AI.

Darktrace has been perfecting behavioral anomaly detection for over a decade; the same self-learning approach that catches BEC and account takeover applies directly to prompt injection delivery. Our multi-layered AI stack extracts content from the message, builds behavioural understanding through social graphing and Pattern of Life analysis, and then combines natural language, topic, inducement, sender relationship and anomaly signals before deciding what action to take.  

This matters for prompt injection because the threat is not the plain language itself, but the intent behind the language that can cause an AI agent to respond in unexpected ways.

How to secure enterprise AI operations from prompt injection attacks

Email was the entry point in this case, but it is only one of many possible vectors.  

Anywhere an agent can retrieve information, an attacker can potentially introduce a prompt injection.

Emails, documents, SharePoint sites, web pages, knowledge bases, chat platforms, and third-party integrations all provide opportunities to influence an agent's behavior. Wherever an agent finds its orders, a prompt injection opportunity exists.

This is why securing AI requires more than blocking malicious inputs. Organizations also need visibility into how agents behave after consuming information from across their environment. If an agent begins accessing unexpected data, taking unusual actions, or operating outside its normal patterns, those behaviors may provide the strongest signal that something has gone wrong.

Effective AI security requires defense in depth: reducing the likelihood of malicious instructions reaching the agent while maintaining the ability to detect and investigate suspicious behavior if they do.

The challenge isn't protecting a single entry point. It's recognizing that, in an AI-powered environment, every source of information is also a potential source of influence.

Are you deploying autonomous agents across your enterprise and want to see this tested in your environment? Let's talk.

Continue reading
About the author
Carlo Loregian
Solutions Engineer
Your data. Our AI.
Elevate your network security with Darktrace AI