See how AI can assist human security teams and think logically to manage cyber incidents efficiently in situations where variables are fast-moving. Read more!
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
Hanah Darley
Director of Threat Research
Share
19
Sep 2023
Within cyber security, crises are a regular occurrence. Whether due to the ever-changing tactics of threat actors or the emergence of new vulnerabilities, security teams find themselves under significant pressure and frequently find themselves in what psychologists term "crisis states."1
A crisis state refers to an internal state marked by confusion and anxiety to such an extent that previously effective coping mechanisms give way to ineffective decision-making and behaviors.2
Given the prevalence of crises in the field of cyber security, practitioners are more prone to consistently making illogical choices due to the intense pressure they experience. They also grapple with a constant influx of rapidly changing information, the need for swift decision-making, and the severe consequences of errors in judgment. They are often asked to assess hundreds of variables and uncertain factors.
The frequency of crisis states is expected to rise as generative AI empowers cyber criminals to accelerate the speed, scale, and sophistication of their attacks.
Why is it so challenging to operate effectively and efficiently during a crisis state? Several factors come into play.
Firstly, individuals are inclined to rely on their instincts, rendering them susceptible to cognitive biases. This makes it increasingly difficult to assimilate new information, process it appropriately, and arrive at logical decisions. Since crises strike unexpectedly and escalate rapidly into new unknowns, responders experience heightened stress, doubt and insecurity when deciding on a course of action.
These cognitive biases manifest in various forms. For instance, confirmation bias prompts people to seek out information that aligns with their pre-existing beliefs, while hindsight bias makes past events seem more predictable in light of present context and information.
Crises also have a profound impact on information processing and decision-making. People tend to simplify new information and often cling to the initial information they receive rather than opting for the most rational decision.
For instance, if an organization has successfully thwarted a ransomware attack in the past, a defender might assume that employing the same countermeasures will suffice for a subsequent attack. However, ransomware tactics are constantly evolving, and a subsequent attack could employ different strategies that evade the previous defenses. In a crisis state, individuals may revert to their prior strategy instead of adapting based on the latest information.
Given there are deeply embedded psychological tendencies and hard-wired decision-making processes leading to a reduction in logic during a crisis, humans need support from technology that does not suffer from the same limitations, particularly in the post-incident phase, where stress levels go into overdrive.
In the era of rapidly evolving novel attacks, security teams require a different approach: AI.
AI can serve as a valuable tool to augment human decision-making, from detection to incident response and mitigation. This is precisely why Darktrace introduced HEAL, which leverages self-learning AI to assist teams in increasing their cyber resilience and managing live incidents, helping to alleviate the cognitive burden they face.
Darktrace HEAL™ learns from your environment, including data points from real incidents and generates simulations to identify the most effective approach for remediation and restoring normal operations. This reduces the overwhelming influx of information and facilitates more effective decision-making during critical moments.
Furthermore, HEAL offers security teams the opportunity to safely simulate realistic attacks within their own environment. Using specific data points from the native environment, simulated incidents prepare security teams for a variety of circumstances which can be reviewed on a regular basis to encourage effective habit forming and reduce cognitive biases from a one-size-fits-all approach. This allows them to anticipate how attacks might unfold and better prepare themselves psychologically for potential real-world incidents.
With the right models and data, AI can significantly mitigate human bias by providing remediation recommendations grounded in evidence and providing proportionate responses based on empirical evidence rather than personal interpretations or instincts. It can act as a guiding light through the chaos of an attack, providing essential support to human security teams.
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.
Out of Character: Detecting Vendor Compromise and Trusted Relationship Abuse with Darktrace
What is Vendor Email Compromise?
Vendor Email Compromise (VEC) refers to an attack where actors breach a third-party provider to exploit their access, relationships, or systems for malicious purposes. The initially compromised entities are often the target’s existing partners, though this can extend to any organization or individual the target is likely to trust.
Itsits at the intersection of supply chain attacks and business email compromise (BEC), blending technical exploitation with trust-based deception. Attackers often infiltrate existing conversations, leveraging AI to mimic tone and avoid common spelling and grammar pitfalls. Malicious content is typically hosted on otherwise reputable file sharing platforms, meaning any shared links initially seem harmless.
While techniques to achieve initial access may have evolved, the goals remain familiar. Threat actors harvest credentials, launch subsequent phishing campaigns, attempt to redirect invoice payments for financial gain, and exfiltrate sensitive corporate data.
Why traditional defenses fall short
These subtle and sophisticated email attacks pose unique challenges for defenders. Few busy people would treat an ongoing conversation with a trusted contact with the same level of suspicion as an email from the CEO requesting ‘URGENT ASSISTANCE!’ Unfortunately, many traditional secure email gateways (SEGs) struggle with this too. Detecting an out-of-character email, when it does not obviously appear out of character, is a complex challenge. It’s hardly surprising, then, that 83% of organizations have experienced a security incident involving third-party vendors [1].
This article explores how Darktrace detected four different vendor compromise campaigns for a single customer, within a two-week period in 2025. Darktrace / EMAIL successfully identified the subtle indicators that these seemingly benign emails from trusted senders were, in fact, malicious. Due to the configuration of Darktrace / EMAIL in this customer’s environment, it was unable to take action against the malicious emails. However, if fully enabled to take Autonomous Response, it would have held all offending emails identified.
How does Darktrace detect vendor compromise?
The answer lies at the core of how Darktrace operates: anomaly detection. Rather than relying on known malicious rules or signatures, Darktrace learns what ‘normal’ looks like for an environment, then looks for anomalies across a wide range of metrics. Despite the resourcefulness of the threat actors involved in this case, Darktrace identified many anomalies across these campaigns.
Different campaigns, common traits
A wide variety of approaches was observed. Individuals, shared mailboxes and external contractors were all targeted. Two emails originated from compromised current vendors, while two came from unknown compromised organizations - one in an associated industry. The sender organizations were either familiar or, at the very least, professional in appearance, with no unusual alphanumeric strings or suspicious top-level domains (TLDs). Subject line, such as “New Approved Statement From [REDACTED]” and “[REDACTED] - Proposal Document” appeared unremarkable and were not designed to provoke heightened emotions like typical social engineering or BEC attempts.
All emails had been given a Microsoft Spam Confidence Level of 1, indicating Microsoft did not consider them to be spam or malicious [2]. They also passed authentication checks (including SPF, and in some cases DKIM and DMARC), meaning they appeared to originate from an authentic source for the sender domain and had not been tampered with in transit.
All observed phishing emails contained a link hosted on a legitimate and commonly used file-sharing site. These sites were often convincingly themed, frequently featuring the name of a trusted vendor either on the page or within the URL, to appear authentic and avoid raising suspicion. However, these links served only as the initial step in a more complex, multi-stage phishing process.
Figure 1: A legitimate file sharing site used in phishing emails to host a secondary malicious link.
Figure 2: Another example of a legitimate file sharing endpoint sent in a phishing email and used to host a malicious link.
If followed, the recipient would be redirected, sometimes via CAPTCHA, to fake Microsoft login pages designed to capturing credentials, namely http://pub-ac94c05b39aa4f75ad1df88d384932b8.r2[.]dev/offline[.]html and https://s3.us-east-1.amazonaws[.]com/s3cure0line-0365cql0.19db86c3-b2b9-44cc-b339-36da233a3be2ml0qin/s3cccql0.19db86c3-b2b9-44cc-b339-36da233a3be2%26l0qn[.]html#.
The latter made use of homoglyphs to deceive the user, with a link referencing ‘s3cure0line’, rather than ‘secureonline’. Post-incident investigation using open-source intelligence (OSINT) confirmed that the domains were linked to malicious phishing endpoints [3] [4].
Figure 3: Fake Microsoft login page designed to harvest credentials.
Figure 4: Phishing kit with likely AI-generated image, designed to harvest user credentials. The URL uses ‘s3cure0line’ instead of ‘secureonline’, a subtle misspelling intended to deceive users.
Darktrace Anomaly Detection
Some senders were unknown to the network, with no previous outbound or inbound emails. Some had sent the email to multiple undisclosed recipients using BCC, an unusual behavior for a new sender.
Where the sender organization was an existing vendor, Darktrace recognized out-of-character behavior, in this case it was the first time a link to a particular file-sharing site had been shared. Often the links themselves exhibited anomalies, either being unusually prominent or hidden altogether - masked by text or a clickable image.
Crucially, Darktrace / EMAIL is able to identify malicious links at the time of processing the emails, without needing to visit the URLs or analyze the destination endpoints, meaning even the most convincing phishing pages cannot evade detection – meaning even the most convincing phishing emails cannot evade detection. This sets it apart from many competitors who rely on crawling the endpoints present in emails. This, among other things, risks disruption to user experience, such as unsubscribing them from emails, for instance.
Darktrace was also able to determine that the malicious emails originated from a compromised mailbox, using a series of behavioral and contextual metrics to make the identification. Upon analysis of the emails, Darktrace autonomously assigned several contextual tags to highlight their concerning elements, indicating that the messages contained phishing links, were likely sent from a compromised account, and originated from a known correspondent exhibiting out-of-character behavior.
Figure 5: Tags assigned to offending emails by Darktrace / EMAIL.
Figure 6: A summary of the anomalous email, confirming that it contained a highly suspicious link.
Out-of-character behavior caught in real-time
In another customer environment around the same time Darktrace / EMAIL detected multiple emails with carefully crafted, contextually appropriate subject lines sent from an established correspondent being sent to 30 different recipients. In many cases, the attacker hijacked existing threads and inserted their malicious emails into an ongoing conversation in an effort to blend in and avoid detection. As in the previous, the attacker leveraged a well-known service, this time ClickFunnels, to host a document containing another malicious link. Once again, they were assigned a Microsoft Spam Confidence Level of 1, indicating that they were not considered malicious.
Figure 7: The legitimate ClickFunnels page used to host a malicious phishing link.
This time, however, the customer had Darktrace / EMAIL fully enabled to take Autonomous Response against suspicious emails. As a result, when Darktrace detected the out-of-character behavior, specifically, the sharing of a link to a previously unused file-sharing domain, and identified the likely malicious intent of the message, it held the email, preventing it from reaching recipients’ inboxes and effectively shutting down the attack.
Figure 8: Darktrace / EMAIL’s detection of malicious emails inserted into an existing thread.*
*To preserve anonymity, all real customer names, email addresses, and other identifying details have been redacted and replaced with fictitious placeholders.
Legitimate messages in the conversation were assigned an Anomaly Score of 0, while the newly inserted malicious emails identified and were flagged with the maximum score of 100.
Key takeaways for defenders
Phishing remains big business, and as the landscape evolves, today’s campaigns often look very different from earlier versions. As with network-based attacks, threat actors are increasingly leveraging legitimate tools and exploiting trusted relationships to carry out their malicious goals, often staying under the radar of security teams and traditional email defenses.
As attackers continue to exploit trusted relationships between organizations and their third-party associates, security teams must remain vigilant to unexpected or suspicious email activity. Protecting the digital estate requires an email solution capable of identifying malicious characteristics, even when they originate from otherwise trusted senders.
Credit to Jennifer Beckett (Cyber Analyst), Patrick Anjos (Senior Cyber Analyst), Ryan Traill (Analyst Content Lead), Kiri Addison (Director of Product)
Appendices
IoC - Type - Description + Confidence
- http://pub-ac94c05b39aa4f75ad1df88d384932b8.r2[.]dev/offline[.]html#p – fake Microsoft login page
- https://s3.us-east-1.amazonaws[.]com/s3cure0line-0365cql0.19db86c3-b2b9-44cc-b339-36da233a3be2ml0qin/s3cccql0.19db86c3-b2b9-44cc-b339-36da233a3be2%26l0qn[.]html# - link to domain used in homoglyph attack
The content provided in this blog is published by Darktrace for general informational purposes only and reflects our understanding of cybersecurity topics, trends, incidents, and developments at the time of publication. While we strive to ensure accuracy and relevance, the information is provided “as is” without any representations or warranties, express or implied. Darktrace makes no guarantees regarding the completeness, accuracy, reliability, or timeliness of any information presented and expressly disclaims all warranties.
Nothing in this blog constitutes legal, technical, or professional advice, and readers should consult qualified professionals before acting on any information contained herein. Any references to third-party organizations, technologies, threat actors, or incidents are for informational purposes only and do not imply affiliation, endorsement, or recommendation.
Darktrace, its affiliates, employees, or agents shall not be held liable for any loss, damage, or harm arising from the use of or reliance on the information in this blog.
The cybersecurity landscape evolves rapidly, and blog content may become outdated or superseded. We reserve the right to update, modify, or remove any content
Announcing Unified OT Security with Dedicated OT Workflows, Segmentation-Aware Risk Insights, and Next-Gen Endpoint Visibility for Industrial Teams
The challenge of convergence without clarity
Convergence is no longer a roadmap idea, it is the daily reality for industrial security teams. As Information Technology (IT) and Operational Technology (OT) environments merge, the line between a cyber incident and an operational disruption grows increasingly hard to define. A misconfigured firewall rule can lead to downtime. A protocol misuse might look like a glitch. And when a pump stalls but nothing appears in the Security Operations Center (SOC) dashboard, teams are left asking: is this operational or is this a threat?
The lack of shared context slows down response, creates friction between SOC analysts and plant engineers, and leaves organizations vulnerable at exactly the points where IT and OT converge. Defenders need more than alerts, they need clarity that both sides can trust.
The breakthrough with Darktrace / OT
This latest Darktrace / OT release was built to deliver exactly that. It introduces shared context between Security, IT, and OT operations, helping reduce friction and close the security gaps at the intersection of these domains.
With a dedicated dashboard built for operations teams, extended visibility into endpoints for new forms of detection and CVE collection, expanded protocol coverage, and smarter risk modeling aligned to segmentation policies, teams can now operate from a shared source of truth. These enhancements are not just incremental upgrades, they are foundational improvements designed to bring clarity, efficiency, and trust to converged environments.
A dashboard built for OT engineers
The new Operational Overview provides OT engineers with a workspace designed for them, not for SOC analysts. It brings asset management, risk insights and operational alerts into one place. Engineers can now see activity like firmware changes, controller reprograms or the sudden appearance of a new workstation on the network, providing a tailored view for critical insights and productivity gains without navigating IT-centric workflows. Each device view is now enriched with cross-linked intelligence, make, model, firmware version and the roles inferred by Self-Learning AI, making it easier to understand how each asset behaves, what function it serves, and where it fits within the broader industrial process. By suppressing IT-centric noise, the dashboard highlights only the anomalies that matter to operations, accelerating triage, enabling smoother IT/OT collaboration, and reducing time to root cause without jumping between tools.
This is usability with purpose, a view that matches OT workflows and accelerates response.
Figure 1: The Operational Overview provides an intuitive dashboard summarizing all OT Assets, Alerts, and Risk.
Full-spectrum coverage across endpoints, sensors and protocols
The release also extends visibility into areas that have traditionally been blind spots. Engineering workstations, Human-Machine Interfaces (HMIs), contractor laptops and field devices are often the entry points for attackers, yet the hardest to monitor.
Darktrace introduces Network Endpoint eXtended Telemetry (NEXT) for OT, a lightweight collector built for segmented and resource-constrained environments. NEXT for OT uses Endpoint sensors to capture localized network, and now process-level telemetry, placing it in context alongside other network and asset data to:
Identify vulnerabilities and OS data, which is leveraged by OT Risk Management for risk scoring and patching prioritization, removing the need for third-party CVE collection.
Surface novel threats using Self-Learning AI that standalone Endpoint Detection and Response (EDR) would miss.
Extend Cyber AI Analyst investigations through to the endpoint root cause.
NEXT is part of our existing cSensor endpoint agent, can be deployed standalone or alongside existing EDR tools, and allows capabilities to be enabled or disabled depending on factors such as security or OT team objectives and resource utilization.
Figure 2: Darktrace / OT delivers CVE patch priority insights by combining threat intelligence with extended network and endpoint telemetry
The family of Darktrace Endpoint sensors also receive a boost in deployment flexibility, with on-prem server-based setups, as well as a Windows driver tailored for zero-trust and high-security environments.
Protocol coverage has been extended where it matters most. Darktrace now performs protocol analysis of a wider range of GE and Mitsubishi protocols, giving operators real-time visibility into commands and state changes on Programmable Logic Controllers (PLCs), robots and controllers. Backed by Self-Learning AI, this inspection does more than parse traffic, it understands what normal looks like and flags deviations that signal risk.
Integrated risk and governance workflows
Security data is only valuable when it drives action. Darktrace / OT delivers risk insights that go beyond patching, helping teams take meaningful steps even when remediation isn't possible. Risk is assessed not just by CVE presence, but by how network segmentation, firewall policies, and attack path logic neutralize or contain real-world exposure. This approach empowers defenders to deprioritize low-impact vulnerabilities and focus effort where risk truly exists. Building on the foundation introduced in release 6.3, such as KEV enrichment, endpoint OS data, and exploit mapping, this release introduces new integrations that bring Darktrace / OT intelligence directly into governance workflows.
Fortinet FortiGate firewall ingestion feeds segmentation rules into attack path modeling, revealing real exposure when policies fail and closing feeds into patching prioritization based on a policy to CVE exposure assessment.
ServiceNow Configuration Management Database (CMDB) sync ensures asset intelligence stays current across governance platforms, eliminating manual inventory work.
Risk modeling has also been made more operationally relevant. Scores are now contextualized by exploitability, asset criticality, firewall policy, and segmentation posture. Patch recommendations are modeled in terms of safety, uptime and compliance rather than just Common Vulnerability Scoring System (CVSS) numbers. And importantly, risk is prioritized across the Purdue Model, giving defenders visibility into whether vulnerabilities remain isolated to IT or extend into OT-critical layers.
Figure 3: Attack Path Modeling based on NetFlow and network topology reveals high risk points of IT/OT convergence.
The real-world impact for defenders
In today’s environments, attackers move fluidly between IT and OT. Without unified visibility and shared context, incidents cascade faster than teams can respond.
With this release, Darktrace / OT changes that reality. The Operational Overview gives Engineers a dashboard they can use daily, tailored to their workflows. SOC analysts can seamlessly investigate telemetry across endpoints, sensors and protocols that were once blind spots. Operators gain transparency into PLCs and controllers. Governance teams benefit from automated integrations with platforms like Fortinet and ServiceNow. And all stakeholders work from risk models that reflect what truly matters: safety, uptime and compliance.
This release is not about creating more alerts. It is about providing more clarity. By unifying context across IT and OT, Darktrace / OT enables defenders to see more, understand more and act faster.
Because in environments where safety and uptime are non-negotiable, clarity is what matters most.
%20Date08132025%200740%20AM.png)
