Blog
/
/
May 5, 2020

The Ongoing Threat of Dharma Ransomware Attacks

Stay informed about the dangers of Dharma ransomware and its methods of attack, ensuring your defenses are strong against potential intrusions.
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
Max Heinemeyer
Global Field CISO
Default blog image
05
May 2020

Executive summary

  • In the past few weeks, Darktrace has observed an increase in attacks against internet-facing systems, such as RDP. The initial intrusions usually take place via existing vulnerabilities or stolen, legitimate credentials. The Dharma ransomware attack described in this blog post is one such example.
  • Old threats can be damaging – Dharma and its variants have been around for four years. This is a classic example of ‘legacy’ ransomware morphing and adapting to bypass traditional defenses.
  • The intrusion shows signs that indicate the threat-actors are aware of – and are actively exploiting – the COVID-19 situation.
  • In the current threat landscape surrounding COVID-19, Darktrace recommends monitoring internet-facing systems and critical servers closely – keeping track of administrative credentials and carefully considering security when rapidly deploying internet-facing infrastructure.

Introduction

In mid-April, Darktrace detected a targeted Dharma ransomware attack on a UK company. The initial point of intrusion was via RDP – this represents a very common attack method of infection that Darktrace has observed in the broader threat landscape over the past few weeks.

This blog post highlights every stage of the attack lifecycle and details the attacker’s techniques, tools and procedures (TTP) – all detected by Darktrace.

Dharma – a varient of the CrySIS malware family – first appeared in 2016 and uses multiple intrusion vectors. It distributes its malware as an attachment in a spam email, by disguising it as an installation file for legitimate software, or by exploiting an open RDP connection through internet-facing servers. When Dharma has finished encrypting files, it drops a ransom note with the contact email address in the encrypted SMB files.

Darktrace had strong, real-time detections of the attack – however the absence of eyes on the user interface prior to the encryption activity, and without Autonomous Response deployed in Active Mode, these alerts were only actioned after the ransomware was unleashed. Fortunately, it was unable to spread within the organization, thanks to human intervention at the peak of the attack. However, Darktrace Antigena in active mode would have significantly slowed down the attack.

Timeline

The timeline below provides a rough overview of the major attack phases over five days of activity.

Figure 1: A timeline of the attack

Technical analysis

Darktrace detected that the main device hit by the attack was an internet-facing RDP server (‘RDP server’). Dharma used network-level encryption here: the ransomware activity takes place over the network protocol SMB.

Below is a chronological overview of all Darktrace detections that fired during this attack: Darktrace detected and reported every single unusual or suspicious event occurring on the RDP server.

Figure 2: An overview of Darktrace detections

Initial compromise

On April 7, the RDP server began receiving a large number of incoming connections from rare IP addresses on the internet.

On April 7, the RDP server began receiving a large number of incoming connections from rare IP addresses on the internet. This means a lot of IP addresses on the internet that usually don’t connect to this company started connection attempts over RDP. The top five cookies used to authenticate show that the source IPs were located in Russia, the Netherlands, Korea, the United States, and Germany.

It is highly likely that the RDP credential used in this attack had been compromised prior to the attack – either via common brute-force methods, credential stuffing attacks, or phishing. Indeed, a TTP growing in popularity is to buy RDP credentials on marketplaces and skip to initial access.

Attempted privilege escalation

The following day, the malicious actor abused the SMB version 1 protocol, notorious for always-on null sessions which offer unauthenticated users’ information about the machine – such as password policies, usernames, group names, machine names, user and host SIDs. What followed was very unusual: the server connected externally to a rare IP address located in Morocco.

Next, the attacker attempted a failed SMB session to the external IP over an unusual port. Darktrace detected this activity as highly anomalous, as it had previously learned that SMB is usually not used in this fashion within this organization – and certainly not for external communication over this port.

Figure 3: Darktrace detecting the rare external IP address

Figure 4: The SMB session failure and the rare connection over port 1047

Command and control traffic

As the entire attack occurred over five days, this aligns with a smash-and-grab approach, rather than a highly covert, low-and-slow operation.

Two hours later, the server initiated a large number of anomalous and rare connections to external destinations located in India, China, and Italy – amongst other destinations the server had never communicated with before. The attacker was now attempting to establish persistence and create stronger channels for command and control (C2). As the entire attack occurred over five days, this aligns with a smash-and-grab approach, rather than a highly covert, low-and-slow operation.

Actions on target

Notwithstanding this approach, the malicious actor remained dormant for two days, biding their time until April 10 — a public holiday in the UK — when security teams would be notably less responsive. This pause in activity provides supporting evidence that the attack was human-driven.

Figure 5: The unusual RDP connections detected by Darktrace

The RDP server then began receiving incoming remote desktop connections from 100% rare IP addresses located in the Netherlands, Latvia, and Poland.

Internal reconnaissance

The IP address 85.93.20[.]6, hosted at the time of investigation in Panama, made two connections to the server, using an administrative credential. On April 12, as other inbound RDP connections scanned the network, the volume of data transferred by the RDP server to this IP address spiked. The RDP server never scans the internal network. Darktrace identified this as highly unusual activity.

Figure 6: Darktrace detects the anomalous external data transfer

Lateral movement and payload execution

Finally, on April 12, the attackers executed the Dharma payload at 13:45. The RDP server wrote a number of files over the SMB protocol, appended with a file extension containing a throwaway email account possibly evoking the current COVID-19 pandemic, ‘cov2020@aol[.]com’. The use of string ‘…@aol.com].ROGER’ and presence of a file named ‘FILES ENCRYPTED.txt’ resembles previous Dharma compromises.

Parallel to the encryption activity, the ransomware tried to spread and infect other machines by initiating successful SMB authentications using the same administrator credential seen during the internal reconnaissance. However, the destination devices did not encrypt any files themselves.

It was during the encryption activity that the internal IT staff pulled the plug from the compromised RDP server, thus ending the ransomware activity.

Conclusion

This incident supports the idea that ‘legacy’ ransomware may morph to resurrect itself to exploit vulnerabilities in remote working infrastructure during this pandemic.

Dharma executed here a fast-acting, planned, targeted, ransomware attack. The attackers used off-the-shelf tools (RDP, abusing SMB1 protocol) blurring detection and attribution by blending in with typical administrator activity.

Darktrace detected every stage of the attack without having to depend on threat intelligence or rules and signatures, and the internal security team acted on the malicious activity to prevent further damage.

This incident supports the idea that ‘legacy’ ransomware may morph to resurrect itself to exploit vulnerabilities in remote working infrastructure during this pandemic. Poorly-secured public-facing systems have been rushed out and security is neglected as companies prioritize availability – sacrificing security in the process. Financially-motivated actors weaponize these weak points.

The use of the COVID-related email ‘cov2020@aol[.]com’ during the attack indicates that the threat-actor is aware of and abusing the current global pandemic.

Recent attacks, such as APT41’s exploitation of the Zoho Manage Engine vulnerability last March, show that attacks against internet-facing infrastructure are gaining popularity as the initial intrusion vector. Indeed, as many as 85% of ransomware attacks use RDP as an entry vector. Ensuring that backups are isolated, configurations are hardened, and systems are patched is not enough – real-time detection of every anomalous action can help protect potential victims of ransomware.

Technical Details

Some of the detections on the RDP server:

  • Compliance / Internet Facing RDP server – exposure of critical server to Internet
  • Anomalous Connection / Application Protocol on Uncommon Port – external connections using an unusual port to rare endpoints
  • Device / Large Number of Connections to New Endpoints – indicative of peer-to-peer or scanning activity
  • Compliance / Incoming Remote Desktop – device is remotely controlled from an external source, increased rick of bruteforce
  • Compromise / Ransomware / Suspicious SMB Activity – reading and writing similar volumes of data to remote file shares, indicative of files being overwritten and encrypted
  • Anomalous File / Internal / Additional Extension Appended to SMB File – device is renaming network share files with an added extension, seen during ransomware activity

The graph below shows the timeline of Darktrace detections on the RDP server. The attack lifecycle is clearly observable.

Figure 7: The model breaches occurring over time

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
Max Heinemeyer
Global Field CISO

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