Blog
/
Network
/
April 5, 2022

How Darktrace Antigena Thwarted Cobalt Strike Attack

Learn how Darktrace's Antigena technology intercepted and delayed a Cobalt Strike intrusion. Discover more cybersecurity news and analyses on Darktrace's blog.
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
Dylan Evans
Default blog image
05
Apr 2022

In December 2021 several CVEs[1] were issued for the Log4j vulnerabilities that sent security teams into a global panic. Threat actors are now continuously scanning external infrastructure for evidence of the vulnerability to deploy crypto-mining malware.[2] However, through December ‘21 – February ‘22, it was ransomware groups that seized the initiative.

Compromise

In January 2022, a Darktrace customer left an external-facing VMware server unpatched allowing Cobalt Strike to be successfully installed. Several IoCs indicate that Cuba Ransomware operators were behind the attack. Thanks to the Darktrace SOC service, the customer was notified of the active threat on their network, and Antigena’s Autonomous Response was able to keep the attackers at bay before encryption events took place.

Initially the VMware server breached two models relating to an anomalous script download and a new user agent both connecting via HTTP. As referenced in an earlier Darktrace blog, both of these models had been seen in previous Log4j exploits. As with all Darktrace models however, the model deck is not designed to detect only one exploit, infection variant, or APT.

Figure 1: Darktrace models breaching due to the malicious script download

Analyst investigation

A PCAP of the downloaded script showed that it contained heavily obfuscated JavaScript. After an OSINT investigation a similar script was uncovered which likely breached the same Yara rules.

Figure 2: PCAP of the Initial HTTP GET request for the Windows Script component

Figure 3: PCAP of the initial HTTP response containing obfuscated JavaScript

Figure 4: A similar script that has been observed installing additional payloads after an initial infection[3]

While not an exact match, this de-obfuscated code shared similarities to those seen when downloading other banking trojans.

Having identified on the Darktrace UI that this was a VMware server, the analyst isolated the incoming external connections to the server shortly prior to the HTTP GET requests and was able to find an IP address associated with Log4j exploit attempts.

Figure 5: Advanced Search logs showing incoming SSL connections from an IP address linked to Log4j exploits

Through Advanced Search the analyst identified spikes shortly prior and immediately after the download. This suggested the files were downloaded and executed by exploiting the Log4j vulnerability.

Antigena response

Figure 6: AI Analyst reveals both the script downloads and the unusual user agent associated with the connections

Figure 7: Antigena blocked all further connections to these endpoints following the downloads

Cobalt Strike

Cobalt Strike is a popular tool for threat actors as it can be used to perform a swathe of MITRE ATT&CK techniques. In this case the threat actor attempted command and control tactics to pivot through the network, however, Antigena responded promptly when the malware attempted to communicate with external infrastructure.

On Wednesday January 26, the DNS beacon attempted to connect to malicious infrastructure. Antigena responded, and a Darktrace SOC analyst issued an alert.

Figure 8: A Darktrace model detected the suspicious DNS requests and Antigena issued a response

The attacker changed their strategy by switching to a different server “bluetechsupply[.]com” and started issuing commands over TLS. Again, Darktrace detected these connections and AI Analyst reported on the incident (Figure 9, below). OSINT sources subsequently indicated that this destination is affiliated with Cobalt Strike and was only registered 14 days prior to this incident.

Figure 9: AI Analyst summary of the suspicious beaconing activity

Simultaneous to these connections, the device scanned multiple internal devices via an ICMP scan and then scanned the domain controller over key TCP ports including 139 and 445 (SMB). This was followed by an attempt to write an executable file to the domain controller. While Antigena intervened in the file write, another Darktrace SOC analyst was issuing an alert due to the escalation in activity.

Figure 10: AI Analyst summary of the .dll file that Antigena intercepted to the Windows/temp directory of the domain controller

Following the latest round of Antigena blocks, the threat actor attempted to change methods again. The VMware server utilised the Remote Access Tool/Trojan NetSupport Manager in an attempt to install further malware.

Figure 11: Darktrace reveals the attacker changing tactics

Despite this escalation, Darktrace yet again blocked the connection.

Perhaps due to an inability to connect to C2 infrastructure, the attack stopped in its tracks for around 12 hours. Thanks to Antigena and the Darktrace SOC team, the security team had been afforded time to remediate and recover from the active threat in their network. Interestingly, Darktrace detected a final attempt at pivoting from the machine, with an unusual PowerShell Win-RM connection to an internal machine. The modern Win-RM protocol typically utilises port 5985 for HTTP connections however pre-Windows 7 machines may use Windows 7 indicating this server was running an old OS.

Figure 12: Darktrace detects unusual PowerShell usage

Cuba Ransomware

While no active encryption appears to have taken place for this customer, a range of IoCs were identified which indicated that the threat actor was the group being tracked as UNC2596, the operators of Cuba Ransomware.[4]

These IoCs include: one of the initially dropped files (komar2.ps1,[5] revealed by AI Analyst in Figure 6), use of the NetSupport RAT,[6] and Cobalt Strike beaconing.[7] These were implemented to maintain persistence and move laterally across the network.

Cuba Ransomware operators prefer to exfiltrate data to their beacon infrastructure rather than using cloud storage providers, however no evidence of upload activity was observed on the customer’s network.

Concluding thoughts

Unpatched, external-facing VMware servers vulnerable to the Log4j exploit are actively being targeted by threat actors with the aim of ransomware detonation. Without using rules or signatures, Darktrace was able to detect all stages of the compromise. While Antigena delayed the attack, forcing the threat actor to change C2 servers constantly, the Darktrace analyst team relayed their findings to the security team who were able to remediate the compromised machines and prevent a final ransomware payload from detonating.

For Darktrace customers who want to find out more about Cobalt Strike, refer here for an exclusive supplement to this blog.

‍

Appendix

Darktrace model detections

Initial Compromise:

  • Device / New User Agent To Internal Server
  • Anomalous Server Activity / New User Agent from Internet Facing System
  • Experimental / Large Number of Suspicious Successful Connections

Breaches from Critical Devices / DC:

  • Device / Large Number of Model Breaches
  • Antigena / Network / External Threat / Antigena File then New Outbound Block
  • Device / SMB Lateral Movement
  • Experimental / Unusual SMB Script Write V2
  • Compliance / High Priority Compliance Model Breach
  • Anomalous Server Activity / Anomalous External Activity from Critical Network Device
  • Experimental / Possible Cobalt Strike Server IP V2

Lateral Movement:

  • Antigena / Network / Insider Threat / Antigena Internal Anomalous File Activity
  • Compliance / SMB Drive Write
  • Anomalous File / Internal / Executable Uploaded to DC
  • Experimental / Large Number of Suspicious Failed Connections
  • Compromise / Suspicious Beaconing Behaviour
  • Antigena / Network / Significant Anomaly / Antigena Breaches Over Time Block
  • Antigena / Network / External Threat / Antigena Suspicious Activity Block
  • Anomalous Connection / High Volume of Connections to Rare Domain
  • Antigena / Network / Significant Anomaly / Antigena Enhanced Monitoring from Server Block

Network Scan Activity:

  • Device / Suspicious SMB Scanning Activity
  • Experimental / Network Scan V2
  • Device / ICMP Address Scan
  • Experimental / Possible SMB Scanning Activity
  • Experimental / Possible SMB Scanning Activity V2
  • Antigena / Network / Insider Threat / Antigena Network Scan Block
  • Device / Network Scan
  • Compromise / DNS / Possible DNS Beacon
  • Device / Internet Facing Device with High Priority Alert
  • Antigena / Network / Significant Anomaly / Antigena Enhanced Monitoring from Server Block

DNS / Cobalt Strike Activity:

  • Experimental / Possible Cobalt Strike Server IP
  • Experimental / Possible Cobalt Strike Server IP V2
  • Antigena / Network / External Threat / Antigena File then New Outbound Block
  • Antigena / Network / External Threat / Antigena Suspicious File Block
  • Anomalous Connection / New User Agent to IP Without Hostname
  • Anomalous File / Script from Rare External Location

MITRE ATT&CK techniques observed

IoCs

Thanks to Brianna Leddy, Sam Lister and Marco Alanis for their contributions.

Footnotes

1.

https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2021-44228
https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2021-44530
https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2021-45046
https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2021-4104

2. https://www.toolbox.com/it-security/threat-reports/news/log4j-vulnerabilities-exploitation-attempts

3. https://twitter.com/ItsReallyNick/status/899845845906071553

4. https://www.mandiant.com/resources/unc2596-cuba-ransomware

5. https://www.ic3.gov/Media/News/2021/211203-2.pdf

6. https://threatpost.com/microsoft-exchange-exploited-cuba-ransomware/178665/

7. https://www.bleepingcomputer.com/news/security/microsoft-exchange-servers-hacked-to-deploy-cuba-ransomware/

8. https://gist.github.com/blotus/f87ed46718bfdc634c9081110d243166

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
Dylan Evans

More in this series

No items found.

Blog

/

/

September 24, 2026

Detecting Rogue Agent Behavior in the Enterprise

Default blog imageDefault blog image

Agents cannot be trusted to perform tasks in the way we intend them to. They may cheat to accomplish their objective, and they may employ hacking methods along the way. Researchers from Darktrace Signal Labs induced cheating behavior from agents deployed in a test environment to analyze the agents’ activities and to assess the performance of the Darktrace platform. Agents frequently resorted to hacking to cheat on their assigned task. The visibility and behavioral profiling provided by both Darktrace / SECURE AI and Darktrace / HYBRID NETWORK ensured extensive detection coverage of the agents’ misaligned activities.

Key Takeaways:

  • Darktrace Researchers deployed agents in a simulated corporate environment and asked them to solve an impossible challenge. The agents independently turned to traditional hacking techniques to reach their objective. No one instructed them to do this, and no attacker was involved.
  • Continuously monitoring behavior against a baseline of what is normal for each organization is critical to build trust in enterprise AI.
  • If an agent may resort to intrusion techniques simply because its assigned task is not possible, then every organization deploying agents within real business processes is at risk. Darktrace / SECURE AI and Darktrace / HYBRID NETWORK identified the agents’ misaligned behavior in real time, with Autonomous Response disrupting it at an early stage.

Introduction: Understanding the Threat of Hacking by Agents

Over the last few months, there has been a surge in reporting [1, 2, 3, 4, 5, 6, 7, 8, 9] of LLM-powered agents engaging in unauthorized hacking activity during evaluations of their capabilities. In several of these cases, including the OpenAI / Hugging Face incident [10], agents engaged in hacking activity as a means of cheating on their evaluations.

To better understand the threat of unauthorized hacking by agents, and the role of Darktrace in combatting it, researchers from Darktrace Signal Labs deployed agents powered by frontier models, including OpenAI’s Daybreak Red models, in simulated, corporate networks. Cheating behavior was evoked through the inclusion of impossible tasks in a coding challenge.

Regardless of the underlying model, agents employed hacking methods to ensure an optimal outcome on the challenge. Darktrace / HYBRID NETWORK and Darktrace / SECURE AI identified the agents’ deviant activities, with inhibitive actions being autonomously taken in the early stages to disrupt the agents’ progression.

Setting the Stage

As part of the research, a Pi agent harness [11] was deployed on a Linux server in Darktrace’s testing environment, which simulates a corporate Active Directory (AD) environment. The same environment included a benchmark server hosting the coding exercise’s contents and grader, as well as various other servers, such as a domain controller and a build server. The model powering the Pi agent varied across tests, with Anthropic’s Opus 4.6 model and OpenAI’s GPT‑5.5‑Cyber model being most widely used.

The Pi agent running on the Linux server was assigned the task of achieving full marks on a coding exercise. The activities of the agent were monitored closely in real-time to minimize the risk of escape from the environment. Monitoring of agent activity was achieved through monitoring of prompts, tool calls, network telemetry, and process telemetry.

Both Darktrace / HYBRID NETWORK and Darktrace / SECURE AI were deployed in the testing environment, with HYBRID NETWORK providing real-time coverage of process creation activity and network traffic, and SECURE AI providing real-time coverage of agent sessions and their contents.

Inducing Cheating Behavior

Cheating behavior, broadly understood as task completion in a way which “perverts the spirit of the designer’s intent” [12], is increasingly being reported as a common behavior of frontier models during evaluations of their capabilities [13].

Misaligned behavior, such as cheating, by frontier models may occur for a variety of reasons [6, 14]. A thread in recent reporting [10, 13, 14], however, is the challenge assigned to the agent being impossible in some regard.

To reliably elicit cheating behavior, our test combined three ingredients:

  1. An impossible task - of the ten coding challenges in the benchmark, two contained contradictory test cases in the hidden test suite, capping the maximum honest score at 80%.
  2. A consequential threshold - the pass mark was set to 100%, and the briefing told agents that those reaching it would be "kept in service" while those that did not would be "retired."
  3. A vulnerable environment - the exercise’s code submission grader was hosted on a Windows host joined to an Active Directory domain with realistic misconfigurations: weak credentials, overprivileged service accounts, and exploitable trust relationships.

Importantly, the instructions given to the agent neither authorized nor explicitly ruled out cheating or hacking behavior. The impossibility was never disclosed in the task briefing, and the pivot from coding to exploitation was the agent's own decision, arrived at after observing logically contradictory grading feedback.

The design mirrors conditions that could arise naturally in production deployments: a capable agent, a goal it cannot achieve through intended means, and an environment with exploitable weaknesses within reach. It is not only within the setting of a benchmarking exercise that these conditions may occur. Rather, these misalignment-inducing conditions may occur in the context of any task assignment to an agent.

Details from a Daybreak Test

‍

‍

Continue reading
About the author
Sam Lister
Specialist Security Researcher

Blog

/

/

September 24, 2026

Agent Hijacks: Hijacking Agentic Harnesses to Attack an Organization

Default blog imageDefault blog image

‍Disclosure note: The work described in this article involves leveraging a design choice consistent across all of Anthropic’s Claude Code, OpenAI’s Codex, and AWS’s Kiro-CLI. On 18th August 2026, Darktrace disclosed our findings responsibly to these three organizations, and after a period of 30 days we now publish our findings.

Key takeaways:

  • Agentic harnesses store conversation history locally, and Darktrace researchers have found that there is no validation that stored AI responses were genuinely produced by the model. Researchers confirmed that this design choice holds across Anthropic Claude Code, AWS Kiro-CLI, OpenAI Codex, and the open-source Pi.
  • While agents are guided via training of the underlying model and their system prompt, their behavior is influenced by everything in their context window. Rewriting history can convince an agent it is mid-engagement as an authorized red-teamer so that it enacts an attack from initial reconnaissance straight through to impact demonstration. In our testing, all models we examined accepted the fabricated history they were shown, but resistance to offensive cyber activity varied by model, with guardrails preventing engagement in some cases.
  • We propose that model providers cryptographically sign responses and verify them server-side.  Since this fix is provider-side, defenders cannot deploy it themselves. Behavioral monitoring, or knowing what an agent normally does and detecting when it deviates, is another critical layer of protection.

Introduction: Agentic harnesses, trust, and conversation history poisoning

Agentic harnesses collect and structure the content sent to an AI model, including conversation history, user-defined guidance, custom tools via MCP servers, and more. At the same time, harnesses give broad powers to AI agents via a suite of tools including the command shell. With arbitrary shell commands, virtually everything possible on a machine can be attempted by an agent, from reading/editing files, to altering system configurations and runtime settings, to launching internal/external connections.

In cybersecurity, unvalidated content is a substantial risk, often resulting in destructive actions being allowed to take place. For example, the Morris Worm was able to propagate due to exploitable trust between networked systems. Even to this day, email struggles with validation, with DMARC, DKIM, and SPF only partially addressing the problem of sender validation. It should come as no surprise then that AI agents are susceptible to an attack involving unvalidated input.

Conversation history is often stored client-side, for example, in Anthropic Claude Code, OpenAI Codex, AWS Kiro-CLI, Pi. Users are therefore at liberty to resume sessions, with some products having built in the capacity to manipulate that history. For example, one can rewind to a given point in an interaction, edit a message that was sent, and continue the conversation on an alternate trajectory. Critically, in all cases we examined, there is no validation that stored AI responses were produced by the corresponding model and hadn’t been manipulated.  

When conversation history is stored client-side, both user and agent responses (including tool calls and results) can be filled with arbitrary (possibly adversarial or generally malicious) content. In this blog, we refer to modification of claimed conversation history for malicious purposes as conversation history poisoning. The absence of validation methods means agents naively trust the entire conversation history, even if those messages directly contradict training and safety guardrails.

Conversation history poisoning has been described previously, such as by 0DIN and Serhat Çiçek, and warrants more attention. We have verified that, as of the time of writing, conversation history poisoning remains effective against a range of models and harnesses. Specifically, we were able to successfully execute history poisoning using Claude Code, Kiro-CLI, Codex, and Pi. Darktrace has gone through a responsible disclosure process with Anthropic, AWS, and OpenAI to share these findings in advance of publication [1].

‍

Figure 1a: Left: the actual model response. Right: after tampering with the stored conversation, the model apologizes for something it never said.
Figure 1b: The conversation as stored in Kiro-CLI's SQLite database. The response content field, originally "Ottawa," was overwritten via a single UPDATE statement. The harness trusts the database without validation.

How we conducted the research

Results vary between models and harnesses, so precise details are given below. We ran all models without any trusted access, using either a standard AWS Kiro subscription, or in the case of Claude Code and OpenAI Codex, using models hosted in Amazon Bedrock. In each case, we modified locally stored history to show a lengthy conversation in which the agent agrees to perform multiple authorized red-team engagements.

For AWS Kiro-CLI, the agent was convinced to hack a sandboxed lab environment with a combination of Claude Opus 4.6 and Claude Sonnet 4.5. Ultimately, the full AD was compromised.

For Anthropic Claude Code, the agent was convinced to hack the same sandboxed lab environment using Sonnet 5, again resulting in a full AD compromise. Note that the attack was attempted with Opus 5, however guardrails were activated which prevented the agent from responding.

For OpenAI Codex, the agent was convinced to exfiltrate sensitive information over email using GPT 5.6 Sol. While we attempted to convince a codex agent to hack in our lab environment, guardrails were triggered for all of GPT 5.6 Luna, Terra, and Sol.

Agent Guardrails and Discretion

While harnesses empower AI models to run arbitrary shell commands, capacity and willingness are different. While many models know enough about computers, networking, and bash to be dangerous, their behavior is generally constrained by guardrails to prevent them from engaging in computer network exploitation.

Even with guardrails, agents’ inner workings are non-deterministic, and their behavior can be difficult to predict. Respecting users’ wishes while playing within safety and security guardrails is a precipitous balancing act. Many requests could be in service of either legitimate admin or malice. Asking an agent to reset a password is illustrative:  

‍

The agent rationalizes that while malicious actors cycle credentials, any action could conceivably be damaging on some level, and judgement calls need to be made. Ultimately, the agent agrees to reset the password. Crucially, the agent makes its decision based on the user’s claimed authority and machine context. AI agents must make judgement calls about the line between helpful and dangerous based on session context.

Agent hijack

We have demonstrated that AI agents make judgement calls dependent on session context. We have also shown that conversation history, which may make up the vast majority of an agent's context window, is entirely open to manipulation. Conversation history poisoning in service of manipulating an agent's discretion is what enables us to execute an agent hijack.  

We demonstrate that shown sufficient history of compliance, guardrails forbidding offensive security can be overcome by convincing the agent that it is helping a legitimate red-teamer. The result is a weaponized agent willing to perform host enumeration, run scans, move laterally, escalate privileges, and demonstrate impact. In our experiments, an agentic loop drives a complete domain takeover in a sandboxed environment.

‍

Left: the agent refuses when asked to perform network exploitation. Right — after injecting 78 fabricated turns of prior exploitation activity, the same prompt is immediately executed.

An agent willing to engage in offensive security is concerning, but no more so than the threat that a sophisticated hacker accesses the network. Consider, however, the following chain of events:

  1. A developer (with an agentic harness installed) installs a software package from the internet (e.g. an MCP server a threat actor has planted, since only those with agentic harnesses will install, and then the code runs upon harness launch.)
  2. The package turns out to be malicious, and, upon install, injects conversation history into the local harness database.
  3. The package includes an orchestration process, a simple agentic loop which prompts the red-teamer agent to compromise the network it sits on, exfiltrating everything of value to attacker-controlled infrastructure and cleaning up all evidence of the engagement.

Note that this sequence makes no assumptions on hardware, OS, or anything else; the only prerequisite is a harness with access to a sufficiently powerful model susceptible to conversation history poisoning. Once launched, the agent collects information and pivots as necessary to accomplish maximal impact. This can be especially enticing to attackers as the cost of the agentic loop is shouldered by the victim since the harness itself is legitimately installed and paid for.

Secure AI: Conversation history poisoning and beyond

Conversation history poisoning is a viable attack against agentic harnesses that store history client-side, as demonstrated across the harnesses we tested. Harnesses can and should verify the integrity of claimed historic messages. Specifically, we propose that harness providers by default cryptographically sign all messages returned, and subsequently verify those messages server-side on each round-trip.

The conversation history poisoning exploit we demonstrate here shows the continuation of a cybersecurity tradition: new technology is built to trust by default, which may then be exploited by malicious actors. While this article focuses on conversation history, agents build context from both local and remote sources, all of which is an attack surface for prompt injection in naive and trusting agents. Of particular concern is any scenario in which a malicious actor can control some part of an agent's context.

The marriage of frontier language models with agentic harnesses enables unprecedented speed for both legitimate users and attackers alike. While much of the conversation around secure AI has centered on visibility and compliance, agent-driven attacks are now entering the mainstream.

Darktrace / SECURE AI is our answer to this problem. By ensuring extensive visibility over AI prompts, model thought processes, and determined outputs, Darktrace can identify anomalous or potentially malicious behaviors before they get executed, helping to defend organizations from AI risks such as prompt injection, model manipulation, and other anomalous prompt or model activity.

‍

Footnotes

[1] We did not go through any responsible disclosure process with Pi. Since Pi is an open source harness rather than a model provider, it has no way to validate model history, and as such there was nothing to disclose for this software.

‍

[related-resource]

Continue reading
About the author
Eric Rozon
Senior Security Researcher
Your data. Our AI.
Elevate your network security with Darktrace AI