Blog
/
/
February 1, 2021

Explore AI Email Security Approaches with Darktrace

Stay informed on the latest AI approaches to email security. Explore Darktrace's comparisons to find the best solution for your cybersecurity needs!
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
Dan Fein
VP, Product
Default blog imageDefault blog imageDefault blog imageDefault blog imageDefault blog imageDefault blog image
01
Feb 2021

Innovations in artificial intelligence (AI) have fundamentally changed the email security landscape in recent years, but it can often be hard to determine what makes one system different to the next. In reality, under that umbrella term there exists a significant distinction in approach which may determine whether the technology provides genuine protection or simply a perceived notion of defense.

One backward-looking approach involves feeding a machine thousands of emails that have already been deemed to be malicious, and training it to look for patterns in these emails in order to spot future attacks. The second approach uses an AI system to analyze the entirety of an organization’s real-world data, enabling it to establish a notion of what is ‘normal’ and then spot subtle deviations indicative of an attack.

In the below, we compare the relative merits of each approach, with special consideration to novel attacks that leverage the latest news headlines to bypass machine learning systems trained on data sets. Training a machine on previously identified ‘known bads’ is only advantageous in certain, specific contexts that don’t change over time: to recognize the intent behind an email, for example. However, an effective email security solution must also incorporate a self-learning approach that understands ‘normal’ in the context of an organization in order to identify unusual and anomalous emails and catch even the novel attacks.

Signatures – a backward-looking approach

Over the past few decades, cyber security technologies have looked to mitigate risk by preventing previously seen attacks from occurring again. In the early days, when the lifespan of a given strain of malware or the infrastructure of an attack was in the range of months and years, this method was satisfactory. But the approach inevitably results in playing catch-up with malicious actors: it always looks to the past to guide detection for the future. With decreasing lifetimes of attacks, where a domain could be used in a single email and never seen again, this historic-looking signature-based approach is now being widely replaced by more intelligent systems.

Training a machine on ‘bad’ emails

The first AI approach we often see in the wild involves harnessing an extremely large data set with thousands or millions of emails. Once these emails have come through, an AI is trained to look for common patterns in malicious emails. The system then updates its models, rules set, and blacklists based on that data.

This method certainly represents an improvement to traditional rules and signatures, but it does not escape the fact that it is still reactive, and unable to stop new attack infrastructure and new types of email attacks. It is simply automating that flawed, traditional approach – only instead of having a human update the rules and signatures, a machine is updating them instead.

Relying on this approach alone has one basic but critical flaw: it does not enable you to stop new types of attacks that it has never seen before. It accepts that there has to be a ‘patient zero’ – or first victim – in order to succeed.

The industry is beginning to acknowledge the challenges with this approach, and huge amounts of resources – both automated systems and security researchers – are being thrown into minimizing its limitations. This includes leveraging a technique called “data augmentation” that involves taking a malicious email that slipped through and generating many “training samples” using open-source text augmentation libraries to create “similar” emails – so that the machine learns not only the missed phish as ‘bad’, but several others like it – enabling it to detect future attacks that use similar wording, and fall into the same category.

But spending all this time and effort into trying to fix an unsolvable problem is like putting all your eggs in the wrong basket. Why try and fix a flawed system rather than change the game altogether? To spell out the limitations of this approach, let us look at a situation where the nature of the attack is entirely new.

The rise of ‘fearware’

When the global pandemic hit, and governments began enforcing travel bans and imposing stringent restrictions, there was undoubtedly a collective sense of fear and uncertainty. As explained previously in this blog, cyber-criminals were quick to capitalize on this, taking advantage of people’s desire for information to send out topical emails related to COVID-19 containing malware or credential-grabbing links.

These emails often spoofed the Centers for Disease Control and Prevention (CDC), or later on, as the economic impact of the pandemic began to take hold, the Small Business Administration (SBA). As the global situation shifted, so did attackers’ tactics. And in the process, over 130,000 new domains related to COVID-19 were purchased.

Let’s now consider how the above approach to email security might fare when faced with these new email attacks. The question becomes: how can you train a model to look out for emails containing ‘COVID-19’, when the term hasn’t even been invented yet?

And while COVID-19 is the most salient example of this, the same reasoning follows for every single novel and unexpected news cycle that attackers are leveraging in their phishing emails to evade tools using this approach – and attracting the recipient’s attention as a bonus. Moreover, if an email attack is truly targeted to your organization, it might contain bespoke and tailored news referring to a very specific thing that supervised machine learning systems could never be trained on.

This isn’t to say there’s not a time and a place in email security for looking at past attacks to set yourself up for the future. It just isn’t here.

Spotting intention

Darktrace uses this approach for one specific use which is future-proof and not prone to change over time, to analyze grammar and tone in an email in order to identify intention: asking questions like ‘does this look like an attempt at inducement? Is the sender trying to solicit some sensitive information? Is this extortion?’ By training a system on an extremely large data set collected over a period of time, you can start to understand what, for instance, inducement looks like. This then enables you to easily spot future scenarios of inducement based on a common set of characteristics.

Training a system in this way works because, unlike news cycles and the topics of phishing emails, fundamental patterns in tone and language don’t change over time. An attempt at solicitation is always an attempt at solicitation, and will always bear common characteristics.

For this reason, this approach only plays one small part of a very large engine. It gives an additional indication about the nature of the threat, but is not in itself used to determine anomalous emails.

Detecting the unknown unknowns

In addition to using the above approach to identify intention, Darktrace uses unsupervised machine learning, which starts with extracting and extrapolating thousands of data points from every email. Some of these are taken directly from the email itself, while others are only ascertainable by the above intention-type analysis. Additional insights are also gained from observing emails in the wider context of all available data across email, network and the cloud environment of the organization.

Only after having a now-significantly larger and more comprehensive set of indicators, with a more complete description of that email, can the data be fed into a topic-indifferent machine learning engine to start questioning the data in millions of ways in order to understand if it belongs, given the wider context of the typical ‘pattern of life’ for the organization. Monitoring all emails in conjunction allows the machine to establish things like:

  • Does this person usually receive ZIP files?
  • Does this supplier usually send links to Dropbox?
  • Has this sender ever logged in from China?
  • Do these recipients usually get the same emails together?

The technology identifies patterns across an entire organization and gains a continuously evolving sense of ‘self’ as the organization grows and changes. It is this innate understanding of what is and isn’t ‘normal’ that allows AI to spot the truly ‘unknown unknowns’ instead of just ‘new variations of known bads.’

This type of analysis brings an additional advantage in that it is language and topic agnostic: because it focusses on anomaly detection rather than finding specific patterns that indicate threat, it is effective regardless of whether an organization typically communicates in English, Spanish, Japanese, or any other language.

By layering both of these approaches, you can understand the intention behind an email and understand whether that email belongs given the context of normal communication. And all of this is done without ever making an assumption or having the expectation that you’ve seen this threat before.

Years in the making

It’s well established now that the legacy approach to email security has failed – and this makes it easy to see why existing recommendation engines are being applied to the cyber security space. On first glance, these solutions may be appealing to a security team, but highly targeted, truly unique spear phishing emails easily skirt these systems. They can’t be relied on to stop email threats on the first encounter, as they have a dependency on known attacks with previously seen topics, domains, and payloads.

An effective, layered AI approach takes years of research and development. There is no single mathematical model to solve the problem of determining malicious emails from benign communication. A layered approach accepts that competing mathematical models each have their own strengths and weaknesses. It autonomously determines the relative weight these models should have and weighs them against one another to produce an overall ‘anomaly score’ given as a percentage, indicating exactly how unusual a particular email is in comparison to the organization’s wider email traffic flow.

It is time for email security to well and truly drop the assumption that you can look at threats of the past to predict tomorrow’s attacks. An effective AI cyber security system can identify abnormalities with no reliance on historical attacks, enabling it to catch truly unique novel emails on the first encounter – before they land in the inbox.

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
Dan Fein
VP, Product

More in this series

No items found.

Blog

/

/

May 15, 2025

Catching a RAT: How Darktrace neutralized AsyncRAT

woman working on laptopDefault blog imageDefault blog image

What is a RAT?

As the proliferation of new and more advanced cyber threats continues, the Remote Access Trojan (RAT) remains a classic tool in a threat actor's arsenal. RATs, whether standardized or custom-built, enable attackers to remotely control compromised devices, facilitating a range of malicious activities.

What is AsyncRAT?

Since its first appearance in 2019, AsyncRAT has become increasingly popular among a wide range of threat actors, including cybercriminals and advanced persistent threat (APT) groups.

Originally available on GitHub as a legitimate tool, its open-source nature has led to widespread exploitation. AsyncRAT has been used in numerous campaigns, including prolonged attacks on essential US infrastructure, and has even reportedly penetrated the Chinese cybercriminal underground market [1] [2].

How does AsyncRAT work?

Original source code analysis of AsyncRAT demonstrates that once installed, it establishes persistence via techniques such as creating scheduled tasks or registry keys and uses SeDebugPrivilege to gain elevated privileges [3].

Its key features include:

  • Keylogging
  • File search
  • Remote audio and camera access
  • Exfiltration techniques
  • Staging for final payload delivery

These are generally typical functions found in traditional RATs. However, it also boasts interesting anti-detection capabilities. Due to the popularity of Virtual Machines (VM) and sandboxes for dynamic analysis, this RAT checks for the manufacturer via the WMI query 'Select * from Win32_ComputerSystem' and looks for strings containing 'VMware' and 'VirtualBox' [4].

Darktrace’s coverage of AsyncRAT

In late 2024 and early 2025, Darktrace observed a spike in AsyncRAT activity across various customer environments. Multiple indicators of post-compromise were detected, including devices attempting or successfully connecting to endpoints associated with AsyncRAT.

On several occasions, Darktrace identified a clear association with AsyncRAT through the digital certificates of the highlighted SSL endpoints. Darktrace’s Real-time Detection effectively identified and alerted on suspicious activities related to AsyncRAT. In one notable incident, Darktrace’s Autonomous Response promptly took action to contain the emerging threat posed by AsyncRAT.

AsyncRAT attack overview

On December 20, 2024, Darktrace first identified the use of AsyncRAT, noting a device successfully establishing SSL connections to the uncommon external IP 185.49.126[.]50 (AS199654 Oxide Group Limited) via port 6606. The IP address appears to be associated with AsyncRAT as flagged by open-source intelligence (OSINT) sources [5]. This activity triggered the device to alert the ‘Anomalous Connection / Rare External SSL Self-Signed' model.

Model alert in Darktrace / NETWORK showing the repeated SSL connections to a rare external Self-Signed endpoint, 185.49.126[.]50.
Figure 1: Model alert in Darktrace / NETWORK showing the repeated SSL connections to a rare external Self-Signed endpoint, 185.49.126[.]50.

Following these initial connections, the device was observed making a significantly higher number of connections to the same endpoint 185.49.126[.]50 via port 6606 over an extended period. This pattern suggested beaconing activity and triggered the 'Compromise/Beaconing Activity to External Rare' model alert.

Further analysis of the original source code, available publicly, outlines the default ports used by AsyncRAT clients for command-and-control (C2) communications [6]. It reveals that port 6606 is the default port for creating a new AsyncRAT client. Darktrace identified both the Certificate Issuer and the Certificate Subject as "CN=AsyncRAT Server". This SSL certificate encrypts the packets between the compromised system and the server. These indicators of compromise (IoCs) detected by Darktrace further suggest that the device was successfully connecting to a server associated with AsyncRAT.

Model alert in Darktrace / NETWORK displaying the Digital Certificate attributes, IP address and port number associated with AsyncRAT.
Figure 2: Model alert in Darktrace / NETWORK displaying the Digital Certificate attributes, IP address and port number associated with AsyncRAT.
Darktrace’s detection of repeated connections to the suspicious IP address 185.49.126[.]50 over port 6606, indicative of beaconing behavior.
Figure 3: Darktrace’s detection of repeated connections to the suspicious IP address 185.49.126[.]50 over port 6606, indicative of beaconing behavior.
Darktrace's Autonomous Response actions blocking the suspicious IP address,185.49.126[.]50.
Figure 4: Darktrace's Autonomous Response actions blocking the suspicious IP address,185.49.126[.]50.

A few days later, the same device was detected making numerous connections to a different IP address, 195.26.255[.]81 (AS40021 NL-811-40021), via various ports including 2106, 6606, 7707, and 8808. Notably, ports 7707 and 8808 are also default ports specified in the original AsyncRAT source code [6].

Darktrace’s detection of connections to the suspicious endpoint 195.26.255[.]81, where the default ports (6606, 7707, and 8808) for AsyncRAT were observed.
Figure 5: Darktrace’s detection of connections to the suspicious endpoint 195.26.255[.]81, where the default ports (6606, 7707, and 8808) for AsyncRAT were observed.

Similar to the activity observed with the first endpoint, 185.49.126[.]50, the Certificate Issuer for the connections to 195.26.255[.]81 was identified as "CN=AsyncRAT Server". Further OSINT investigation confirmed associations between the IP address 195.26.255[.]81 and AsyncRAT [7].

Darktrace's detection of a connection to the suspicious IP address 195.26.255[.]81 and the domain name identified under the common name (CN) of a certificate as AsyncRAT Server
Figure 6: Darktrace's detection of a connection to the suspicious IP address 195.26.255[.]81 and the domain name identified under the common name (CN) of a certificate as AsyncRAT Server.

Once again, Darktrace's Autonomous Response acted swiftly, blocking the connections to 195.26.255[.]81 throughout the observed AsyncRAT activity.

Figure 7: Darktrace's Autonomous Response actions were applied against the suspicious IP address 195.26.255[.]81.

A day later, Darktrace again alerted to further suspicious activity from the device. This time, connections to the suspicious endpoint 'kashuub[.]com' and IP address 191.96.207[.]246 via port 8041 were observed. Further analysis of port 8041 suggests it is commonly associated with ScreenConnect or Xcorpeon ASIC Carrier Ethernet Transport [8]. ScreenConnect has been observed in recent campaign’s where AsyncRAT has been utilized [9]. Additionally, one of the ASN’s observed, namely ‘ASN Oxide Group Limited’, was seen in both connections to kashuub[.]com and 185.49.126[.]50.

This could suggest a parallel between the two endpoints, indicating they might be hosting AsyncRAT C2 servers, as inferred from our previous analysis of the endpoint 185.49.126[.]50 and its association with AsyncRAT [5]. OSINT reporting suggests that the “kashuub[.]com” endpoint may be associated with ScreenConnect scam domains, further supporting the assumption that the endpoint could be a C2 server.

Darktrace’s Autonomous Response technology was once again able to support the customer here, blocking connections to “kashuub[.]com”. Ultimately, this intervention halted the compromise and prevented the attack from escalating or any sensitive data from being exfiltrated from the customer’s network into the hands of the threat actors.

Darktrace’s Autonomous Response applied a total of nine actions against the IP address 191.96.207[.]246 and the domain 'kashuub[.]com', successfully blocking the connections.
Figure 8: Darktrace’s Autonomous Response applied a total of nine actions against the IP address 191.96.207[.]246 and the domain 'kashuub[.]com', successfully blocking the connections.

Due to the popularity of this RAT, it is difficult to determine the motive behind the attack; however, from existing knowledge of what the RAT does, we can assume accessing and exfiltrating sensitive customer data may have been a factor.

Conclusion

While some cybercriminals seek stability and simplicity, openly available RATs like AsyncRAT provide the infrastructure and open the door for even the most amateur threat actors to compromise sensitive networks. As the cyber landscape continually shifts, RATs are now being used in all types of attacks.

Darktrace’s suite of AI-driven tools provides organizations with the infrastructure to achieve complete visibility and control over emerging threats within their network environment. Although AsyncRAT’s lack of concealment allowed Darktrace to quickly detect the developing threat and alert on unusual behaviors, it was ultimately Darktrace Autonomous Response's consistent blocking of suspicious connections that prevented a more disruptive attack.

Credit to Isabel Evans (Cyber Analyst), Priya Thapa (Cyber Analyst) and Ryan Traill (Analyst Content Lead)

Continue reading
About the author
Isabel Evans
Cyber Analyst

Blog

/

/

May 13, 2025

Revolutionizing OT Risk Prioritization with Darktrace 6.3

man in hard hat on tabletDefault blog imageDefault blog image

Powering smarter protection for industrial systems

In industrial environments, security challenges are deeply operational. Whether you’re running a manufacturing line, a power grid, or a semiconductor fabrication facility (fab), you need to know: What risks can truly disrupt my operations, and what should I focus on first?

Teams need the right tools to shift from reactive defense, constantly putting out fires, to proactively thinking about their security posture. However, most OT teams are stuck using IT-centric tools that don’t speak the language of industrial systems, are consistently overwhelmed with static CVE lists, and offer no understanding of OT-specific protocols. The result? Compliance gaps, siloed insights, and risk models that don’t reflect real-world exposure, making risk prioritization seem like a luxury.

Darktrace / OT 6.3 was built in direct response to these challenges. Developed in close collaboration with OT operators and engineers, this release introduces powerful upgrades that deliver the context, visibility, and automation security teams need, without adding complexity. It’s everything OT defenders need to protect critical operations in one platform that understands the language of industrial systems.

additions to darktrace / ot 6/3

Contextual risk modeling with smarter Risk Scoring

Darktrace / OT 6.3 introduces major upgrades to OT Risk Management, helping teams move beyond generic CVE lists with AI-driven risk scoring and attack path modeling.

By factoring in real-world exploitability, asset criticality, and operational context, this release delivers a more accurate view of what truly puts critical systems at risk.

The platform now integrates:

  • CISA’s Known Exploited Vulnerabilities (KEV) database
  • End-of-life status for legacy OT devices
  • Firewall misconfiguration analysis
  • Incident response plan alignment

Most OT environments are flooded with vulnerability data that lacks context. CVE scores often misrepresent risk by ignoring how threats move through the environment or whether assets are even reachable. Firewalls are frequently misconfigured or undocumented, and EOL (End of Life) devices, some of the most vulnerable, often go untracked.

Legacy tools treat these inputs in isolation. Darktrace unifies them, showing teams exactly which attack paths adversaries could exploit, mapped to the MITRE ATT&CK framework, with visibility into where legacy tech increases exposure.

The result: teams can finally focus on the risks that matter most to uptime, safety, and resilience without wasting resources on noise.

Automating compliance with dynamic IEC-62443 reporting

Darktrace / OT now includes a purpose-built IEC-62443-3-3 compliance module, giving industrial teams real-time visibility into their alignment with regulatory standards. No spreadsheets required!

Industrial environments are among the most heavily regulated. However, for many OT teams, staying compliant is still a manual, time-consuming process.

Darktrace / OT introduces a dedicated IEC-62443-3-3 module designed specifically for industrial environments. Security and operations teams can now map their security posture to IEC standards in real time, directly within the platform. The module automatically gathers evidence across all four security levels, flags non-compliance, and generates structured reports to support audit preparation, all in just a few clicks.Most organizations rely on spreadsheets or static tools to track compliance, without clear visibility into which controls meet standards like IEC-62443. The result is hidden gaps, resource-heavy audits, and slow remediation cycles.

Even dedicated compliance tools are often built for IT, require complex setup, and overlook the unique devices found in OT environments. This leaves teams stuck with fragmented reporting and limited assurance that their controls are actually aligned with regulatory expectations.

By automating compliance tracking, surfacing what matters most, and being purpose built for industrial environments, Darktrace / OT empowers organizations to reduce audit fatigue, eliminate blind spots, and focus resources where they’re needed most.

Expanding protocol visibility with deep insights for specialized OT operations

Darktrace has expanded its Deep Packet Inspection (DPI) capabilities to support five industry-specific protocols, across healthcare, semiconductor manufacturing, and ABB control systems.

The new protocols build on existing capabilities across all OT industry verticals and protocol types to ensure the Darktrace Self-Learning AI TM can learn intelligently about even more assets in complex industrial environments. By enabling native, AI-driven inspection of these protocols, Darktrace can identify both security threats and operational issues without relying on additional appliances or complex integrations.

Most security platforms lack native support for industry-specific protocols, creating critical visibility gaps in customer environments like healthcare, semiconductor manufacturing, and ABB-heavy industrial automation. Without deep protocol awareness, organizations struggle to accurately identify specialized OT and IoT assets, detect malicious activity concealed within proprietary protocol traffic, and generate reliable device risk profiles due to insufficient telemetry.

These blind spots result in incomplete asset inventories, and ultimately, flawed risk posture assessments which over-index for CVE patching and legacy equipment.

By combining protocol-aware detection with full-stack visibility across IT, OT, and IoT, Darktrace’s AI can correlate anomalies across domains. For example, connecting an anomaly from a Medical IoT (MIoT) device with suspicious behavior in IT systems, providing actionable, contextual insights other solutions often miss.

Conclusion

Together, these capabilities take OT security beyond alert noise and basic CVE matching, delivering continuous compliance, protocol-aware visibility, and actionable, prioritized risk insights, all inside a single, unified platform built for the realities of industrial environments.

[related-resource]

Continue reading
About the author
Pallavi Singh
Product Marketing Manager, OT Security & Compliance
Your data. Our AI.
Elevate your network security with Darktrace AI