Blog
/
Email
/
August 2, 2024

Safelink Smuggling: Enhancing Resilience Against Malicious Links

Gain insights into safelink smuggling tactics and learn strategies to protect your organization from the dangers posed by malicious links.
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
Carlos Gray
Senior Product Marketing Manager, Email
Written by
Stephen Pickman
Default blog imageDefault blog imageDefault blog imageDefault blog imageDefault blog imageDefault blog image
02
Aug 2024

Darktrace security members and researchers have recently seen a rise in what we are calling Safelink Smuggling. Safelinks are URLs rewritten by security solutions to enable additional analysis when the URL is clicked. Once analyzed, they may prompt a user, redirect the browser back to the original URL, or block further access if deemed necessary.

What is Safelink Smuggling?

Safelink Smuggling is a technique that involves an attacker purposely getting their malicious payload rewritten by a security solution’s Safelink capability to then propagate the rewritten URL to others. This technique is a way for attackers to not only avoid detection by traditional email security and other solutions, but also to instill mistrust in all email security solutions. As a result, Safelinks from a range of popular email security providers are often seen in phishing or supply chain attacks. In fact, Darktrace has observed over 300,000 cases of Safelinks being included in unexpected and suspicious contexts over the last 3 months.

How does Safelink Smuggling work?

Safelink Smuggling has two key stages: Getting a malicious link rewritten by an email security solution, then propagating that rewritten link to other victims.

Step one:

Obfuscated a malicious payload through a Safelink capability rewriting the link; Darktrace has seen this attempted through two methods – Compromised Account or Reply-Chain.

  • Method 1: Compromised Account

If an attacker can gain access to a compromised account – whether that’s through brute force, malware or credential theft – they can infiltrate it with malicious links, and then exfiltrate the Safelinks created as the email passes through security filtering. In other words, attackers will send a malicious payload to the compromised inbox, with the intent that the malicious URL gets rewritten. Unlike a normal phishing email where the threat actor wants to avoid having their email blocked, in this case the objective is for the email to get through to the inbox with the link rewritten. As observed by Darktrace, attackers often send the link in isolation as any additional components (i.e., body text or other content in the email) could cause a more severe action such as the email security solution holding the message.

  • Method 2: Reply-Chain

With this method, the attacker sends a malicious link to an email security vendor’s customer in an attempt to solicit a reply from an internal user. This allows them to grab the re-written URL within the reply chain. However, this is a risky tactic which can fail at several points. The attacker has to be confident the initial email won't be blocked outright; they also risk alerting security vendors to the address and the URL intended to be used for the main campaign. They also must be confident that the checks made when the re-written URL is clicked will not lead to a block at the final destination.
Regardless of the method used, the end result will appear as follows:

For example, the original malicious URL may look like this,

faceldu[.]org/Invoice112.zip

(negative surface indicators: recently registered domain, file extension)

And after being rewritten,

securityvevndor[.]com/safe?q=aNDF80dfaAkAH930adbd

(positive surface indicators: established domain, positive reputation, associated with safe content)

Step Two:

Now that the attacker has access to a malicious URL that has been obfuscated by a safe rewrite, attackers can forward or craft an email leveraging that same link. In fact, we have even seen multiple layers of Safelink Smuggling being used to mask a payload further.

The Challenge of Link Rewriting

Traditional email security solutions rewrite all links sent to an organization, but there is an inherent risk to this methodology. Rewriting every link, whether harmless or harmful, leads employees to lose context and creates a false sense of security when interacting with rewritten links in emails. Furthermore, it provides attackers with many opportunities to exploit Safelinks. As demonstrated in Method 2 above, if an email security solution does not rewrite every link, executing such attacks would be significantly more challenging.

Traditionally, rewriting every link made sense from a security perspective, as it allowed servers to thoroughly analyze links for known attack patterns and signatures. However, this approach relies on identifying previously recognized threats. Conversely, Darktrace / EMAIL gathers sufficient information about a link without needing to rewrite it, by analyzing the context and content of the email and the link itself.

In fact, Darktrace is the pioneer in applying selective rewriting to URLs based on suspicious properties or context, a method that other solutions have since adopted. While traditional solutions rewrite links to assess them only after they are clicked, Darktrace / EMAIL takes immediate action to neutralize threats before they reach the inbox.

Darktrace achieves high success rates in detecting malicious links and emails on the first encounter using Self-Learning AI. By understanding 'normal' behavior in email communications, Darktrace identifies subtle deviations indicative of cyber threats and selectively rewrites only those links deemed suspicious, ensuring a targeted, proportionate, and non-disruptive response.

Why do traditional email security solutions miss Safelink attacks?

Traditional security solutions that focus on learning attack patterns will miss Safelink threats as they are often utilized in attacks that have a variety of layers which help the email seem legitimate. Leveraging all the classic techniques seen in a supply chain attack to disguise the sender's intent, taking advantage of the users' inherent trust in familiar sources, the user is more likely to lower their defenses.

For more information: https://darktrace.com/products/email/use-cases/supply-chain-attack

In terms of the URL, if the payload is malicious, why is it difficult for email security solutions to catch it? Primarily, other security vendors will focus on the payload in isolation, attempting to find known attack patterns or signatures such as a domain name or IP with a bad reputation. Unfortunately, with this technique, if the URL has a legitimate domain, it will return a clean track record. Common obfuscation techniques such as captchas, short-links, and click throughs can all be deployed to add layers of complexity to the analysis.

Safelink Smuggling relies heavily on link redirects, which means that web analysis tools will falter as they will only analyze the first redirect. Consequently, when more in-depth analysis on the link itself is performed, the first place the URL takes the user is not the malicious site but rather the default on-click analysis of the vendor in question. Therefore, any traditional browser or link analysis will also return a negative result.

Finally, the context itself is important. In contrast to traditional email security solutions, Darktrace / EMAIL asks who, what, when, where, and why for every single email, and compares it to the pattern of life of both the internal recipient and the external sender, rather than attempting to match patterns with historical threat data. When analyzing an email from an inbound perspective, Darktrace reveals potential deviations from normal, that, when considered sufficiently anomalous, will result in taking a proportional action to the threat assessed.

To illustrate the above, let’s take a look at an example email that Darktrace recently caught.

The following is an email a Darktrace customer received, which Darktrace / EMAIL held before it reached the inbox. In this case, the smuggled Safelink was further obfuscated behind a QR Code. The accompanying document also presented some anomalies in terms of its intent, perceived as a potential social engineering attempt. Finally, the lack of association and low mailing history meant there was no prior context for this email.  

Example of a Safelink Smuggling attack using a popular email security solution’s safelink.
Fig 1: Example of a Safelink Smuggling attack using a popular email security solution’s safelink.

How to mitigate against Safelink Smuggling?

It's difficult for email security vendors to do anything about their links being reused, and reuse should almost be expected by popular operators in the email security space. Therefore, the presence of links from a vendor’s domain in a suspicious email communication rarely indicates a compromise of the link rewrite infrastructure or a compromise of the third-party vendor.

Email security vendors can improve their defense-in-depth, especially around their email provider accounts to avoid Method 1 (Compromised Account attacks) and become more selective with their rewrites to curtail Method 2 (Reply Chain attacks).

Primary protection against Safelink Smuggling should be offered by the email security vendor responsible for inbound email analysis. They need to ensure that techniques such as Safelink Smuggling are not evaded by their detection mechanisms.

Darktrace has long been working on the betterment of security within the email community and innovating our link analysis infrastructure to mitigate against this attack methodology (read more about our major update in 6.2 here), regardless of whether the receiving organization are Darktrace customers.

How does Darktrace deal with Safelink Smuggling today?

Darktrace has been dealing with Safelink Smuggling since launch and has a standardized recommendation for customers who are looking to defend against this threat.

Customers want to avoid being 1) the propagators of this threat and potentially damaging their brand reputation, and 2) being victims of the supply chain attack thereafter.

The principal recommendation to protect customer accounts and consequently their brands is to ensure defense-in-depth. As accounts establish themselves as the crown jewels of any modern enterprise, organizations should vigilantly monitor their account activity with the same rigor they would analyze their network activity. Whether that is through the base account takeover protection offered by Darktrace / EMAIL, or the expanded defense offered by Darktrace / IDENTITY, it is crucial that the accounts themselves have a robust security solution in place.

Secondly, to avoid falling victim to the supply chain attack that leverages a third-party vendor’s link rewrite, it is imperative to use a solution that does not rely on static threat intelligence and link reputation analysis. Rather than chasing attackers by updating rules and signatures, Darktrace leverages Self-Learning AI to learn the communication patterns of both internal and external messages to reveal deviations in both content and context.

Finally, for those customers that already leverage Darktrace / EMAIL we recommend ensuring that lock links are enabled, and that the default warning page is displayed every time a link is rewritten, no matter the perceived severity of the link. This will allow any potential user that clicks on a rewritten Darktrace / EMAIL link to be alerted to the potential nature of the site they are trying to access.

Safelink smuggling example caught by Darktrace

While most cases involve other vendors, analysts recently saw a case where Darktrace's own links were used in this type of attack. A small number of links were leveraged in a campaign targeting both Darktrace and non-Darktrace customers alike. Thankfully, these attempts were all appropriately actioned by those customers that had Darktrace / EMAIL deployed.

In the example below, you will see how Darktrace Cyber AI Analyst describes the example at hand under the Anomaly Indicators section.

Example of Safelink Smuggling attack on Darktrace using the Darktrace Safelink Infrastructure.
Fig 2: Example of Safelink Smuggling attack on Darktrace using the Darktrace Safelink Infrastructure.

First, the display name mismatch can be interpreted as an indicator of social engineering, attempting to deceive the recipient with an IT policy change.

Second, the link itself, which in this case is a hidden redirect to an unusual host for this environment.

Finally, there is a suspected account takeover due to the origin of the email being a long-standing, validated domain that contains a wide variety of suspicious elements.

Darktrace / EMAIL would have held this email from being delivered.

Conclusion

By investigating Safelink Smuggling, Darktrace wants to shine a light on the technique for security teams and help raise awareness of how it can be used to dupe users into lowering their defenses. Challenge your email security vendor on how it deals with link analysis, particularly from trusted senders and applications.

Interested in Darktrace’s approach to defense-in-depth? Check out Darktrace / EMAIL

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
Carlos Gray
Senior Product Marketing Manager, Email
Written by
Stephen Pickman

More in this series

No items found.

Blog

/

Email

/

May 21, 2025

Evaluating Email Security: How to Select the Best Solution for Your Organization

person holding ipadDefault blog imageDefault blog image

When evaluating email security solutions, it’s crucial to move beyond marketing claims and focus on real-world performance. One of the most effective ways to achieve this is through an A/B comparison approach – a side-by-side evaluation of vendors based on consistent, predefined criteria.

This method cuts through biases, reveals true capability differences, and ensures that all solutions are assessed on a level playing field. It’s not just about finding an objectively good solution – it’s about finding the best solution for your organization’s specific needs.

An A/B comparison approach is particularly effective for three main reasons:

  1. Eliminates bias: By comparing solutions under identical conditions, it’s easier to spot differences in performance without the fog of marketing jargon.
  2. Highlights real capabilities: Direct side-by-side testing exposes genuine strengths and weaknesses, making it easier to judge which features are impactful versus merely decorative.
  3. Encourages objective decision-making: This structured method reduces emotional or brand-driven decisions, focusing purely on metrics and performance.

Let’s look at the key factors to consider when setting up your evaluation to ensure a fair, accurate, and actionable comparison.

Deployment: Setting the stage for fair evaluation

To achieve a genuine comparison, deployment must be consistent across all evaluated solutions:

  • Establish the same scope: All solutions should be granted identical visibility across relevant tenants and domains to ensure parity.
  • Set a concrete timeline: Deploy and test each solution with the same dataset, at the same points in time. This allows you to observe differences in learning periods and adaptive capabilities.

Equal visibility and synchronized timelines prevent discrepancies that could skew your understanding of each vendor’s true capabilities. But remember – quicker results might not equal better learning or understanding!

Tuning and configurations: Optimizing for real-world conditions

Properly tuning and configuring each solution is critical for fair evaluation:

  • Compare on optimal performance: Consult with each vendor to understand what optimal deployment looks like for their solution, particularly if machine learning is involved.
  • Consider the long term: Configuration adjustments should be made with long-term usage in mind. Short-term fixes can mask long-term challenges.
  • Data visibility: Ensure each solution can retain and provide search capabilities on all data collected throughout the evaluation period.

These steps guarantee that you are comparing fully optimized versions of each platform, not underperforming or misconfigured ones.

Evaluation: Applying consistent metrics

Once deployment and configurations are aligned, the evaluation itself must be consistent, to prevent unfair scoring and help to identify true differences in threat detection and response capabilities.

  • Coordinate your decision criteria: Ensure all vendors are measured against the same set of criteria, established before testing begins.
  • Understand vendor threat classification: Each vendor may have different ways of classifying threats, so be sure to understand these nuances.
  • Maintain communication: If results seem inaccurate, engage with the vendors. Their response and remediation capabilities are part of the evaluation.

Making a decision: Look beyond the metrics

When it comes to reviewing the performance of each solution, it’s important to both consider and look beyond the raw data. This is about choosing the solution that best aligns with your specific business needs, which may include factors and features not captured in the results.

  • Evaluate based on results: Consider accuracy, threats detected, precision, and response effectiveness.
  • Evaluate beyond results: Assess the overall experience, including support, integrations, training, and long-term alignment with your security strategy.
  • Review and communicate: Internally review the findings and communicate them back to the vendors.

Choosing the right email security solution isn’t just about ticking boxes, it’s about strategic alignment with your organization’s goals and the evolving threat landscape. A structured, A/B comparison approach will help ensure that the solution you select is truly the best fit.

For a full checklist of the features and capabilities to compare, as well as how to perform a commercial and technical evaluation, check out the full Buyer’s Checklist for Evaluating Email Security.

[related-resource]

Continue reading
About the author
Carlos Gray
Senior Product Marketing Manager, Email

Blog

/

OT

/

May 21, 2025

Adapting to new USCG cybersecurity mandates: Darktrace for ports and maritime systems

Cargo ships at a portDefault blog imageDefault blog image

What is the Marine Transportation System (MTS)?

Marine Transportation Systems (MTS) play a substantial roll in U.S. commerce, military readiness, and economic security. Defined as a critical national infrastructure, the MTS encompasses all aspects of maritime transportation from ships and ports to the inland waterways and the rail and roadways that connect them.

MTS interconnected systems include:

  • Waterways: Coastal and inland rivers, shipping channels, and harbors
  • Ports: Terminals, piers, and facilities where cargo and passengers are transferred
  • Vessels: Commercial ships, barges, ferries, and support craft
  • Intermodal Connections: Railroads, highways, and logistics hubs that tie maritime transport into national and global supply chains

The Coast Guard plays a central role in ensuring the safety, security, and efficiency of the MTS, handling over $5.4 trillion in annual economic activity. As digital systems increasingly support operations across the MTS, from crane control to cargo tracking, cybersecurity has become essential to protecting this lifeline of U.S. trade and infrastructure.

Maritime Transportation Systems also enable international trade, making them prime targets for cyber threats from ransomware gangs to nation-state actors.

To defend against growing threats, the United States Coast Guard (USCG) has moved from encouraging cybersecurity best practices to enforcing them, culminating in a new mandate that goes into effect on July 16, 2025. These regulations aim to secure the digital backbone of the maritime industry.

Why maritime ports are at risk

Modern ports are a blend of legacy and modern OT, IoT, and IT digitally connected technologies that enable crane operations, container tracking, terminal storage, logistics, and remote maintenance.

Many of these systems were never designed with cybersecurity in mind, making them vulnerable to lateral movement and disruptive ransomware attack spillover.

The convergence of business IT networks and operational infrastructure further expands the attack surface, especially with the rise of cloud adoption and unmanaged IoT and IIoT devices.

Cyber incidents in recent years have demonstrated how ransomware or malicious activity can halt crane operations, disrupt logistics, and compromise safety at scale threatening not only port operations, but national security and economic stability.

Relevant cyber-attacks on maritime ports

Maersk & Port of Los Angeles (2017 – NotPetya):
A ransomware attack crippled A.P. Moller-Maersk, the world’s largest shipping company. Operations at 17 ports, including the Port of Los Angeles, were halted due to system outages, causing weeks of logistical chaos.

Port of San Diego (2018 – Ransomware Attack):
A ransomware attack targeted the Port of San Diego, disrupting internal IT systems including public records, business services, and dockside cargo operations. While marine traffic was unaffected, commercial activity slowed significantly during recovery.

Port of Houston (2021 – Nation-State Intrusion):
A suspected nation-state actor exploited a known vulnerability in a Port of Houston web application to gain access to its network. While the attack was reportedly thwarted, it triggered a federal investigation and highlighted the vulnerability of maritime systems.

Jawaharlal Nehru Port Trust, India (2022 – Ransomware Incident):
India’s largest container port experienced disruptions due to a ransomware attack affecting operations and logistics systems. Container handling and cargo movement slowed as IT systems were taken offline during recovery efforts.

A regulatory shift: From guidance to enforcement

Since the Maritime Transportation Security Act (MTSA) of 2002, ports have been required to develop and maintain security plans. Cybersecurity formally entered the regulatory fold in 2020 with revisions to 33 CFR Part 105 and 106, requiring port authorities to assess and address computer system vulnerabilities.

In January 2025, the USCG finalized new rules to enforce cybersecurity practices across the MTS. Key elements include (but are not limited to):

  • A dedicated cyber incident response plan (PR.IP-9)
  • Routine cybersecurity risk assessments and exercises (ID.RA)
  • Designation of a cybersecurity officer and regular workforce training (section 3.1)
  • Controls for access management, segmentation, logging, and encryption (PR.AC-1:7)
  • Supply chain risk management (ID.SC)
  • Incident reporting to the National Response Center

Port operators are encouraged to align their programs with the NIST Cybersecurity Framework (CSF 2.0) and NIST SP 800-82r3, which provide comprehensive guidance for IT and OT security in industrial environments.

How Darktrace can support maritime & ports

Unified IT + OT + Cloud coverage

Maritime ports operate in hybrid environments spanning business IT systems (finance, HR, ERP), industrial OT (cranes, gates, pumps, sensors), and an increasing array of cloud and SaaS platforms.

Darktrace is the only vendor that provides native visibility and threat detection across OT/IoT, IT, cloud, and SaaS environments — all in a single platform. This means:

  • Cranes and other physical process control networks are monitored in the same dashboard as Active Directory and Office 365.
  • Threats that start in the cloud (e.g., phishing, SaaS token theft) and pivot or attempt to pivot into OT are caught early — eliminating blind spots that siloed tools miss.

This unification is critical to meeting USCG requirements for network-wide monitoring, risk identification, and incident response.

AI that understands your environment. Not just known threats

Darktrace’s AI doesn’t rely on rules or signatures. Instead, it uses Self-Learning AI TM that builds a unique “pattern of life” for every device, protocol, user, and network segment, whether it’s a crane router or PLC, SCADA server, Workstation, or Linux file server.

  • No predefined baselines or manual training
  • Real-time anomaly detection for zero-days, ransomware, and supply chain compromise
  • Continuous adaptation to new devices, configurations, and operations

This approach is critical in diverse distributed OT environments where change and anomalous activity on the network are more frequent. It also dramatically reduces the time and expertise needed to classify and inventory assets, even for unknown or custom-built systems.

Supporting incident response requirements

A key USCG requirement is that cybersecurity plans must support effective incident response.

Key expectations include:

  • Defined response roles and procedures: Personnel must know what to do and when (RS.CO-1).
  • Timely reporting: Incidents must be reported and categorized according to established criteria (RS.CO-2, RS.AN-4).
  • Effective communication: Information must be shared internally and externally, including voluntary collaboration with law enforcement and industry peers (RS.CO-3 through RS.CO-5).
  • Thorough analysis: Alerts must be investigated, impacts understood, and forensic evidence gathered to support decision-making and recovery (RS.AN-1 through RS.AN-5).
  • Swift mitigation: Incidents must be contained and resolved efficiently, with newly discovered vulnerabilities addressed or documented (RS.MI-1 through RS.MI-3).
  • Ongoing improvement: Organizations must refine their response plans using lessons learned from past incidents (RS.IM-1 and RS.IM-2).

That means detections need to be clear, accurate, and actionable.

Darktrace cuts through the noise using AI that prioritizes only high-confidence incidents and provides natural-language narratives and investigative reports that explain:

  • What’s happening, where it’s happening, when it’s happening
  • Why it’s unusual
  • How to respond

Result: Port security teams often lean and multi-tasked can meet USCG response-time expectations and reporting needs without needing to scale headcount or triage hundreds of alerts.

Built-for-edge deployment

Maritime environments are constrained. Many traditional SaaS deployment types often are unsuitable for tugboats, cranes, or air-gapped terminal systems.

Darktrace builds and maintains its own ruggedized, purpose-built appliances and unique virtual deployment options that:

  • Deploy directly into crane networks or terminal enclosures
  • Require no configuration or tuning, drop-in ready
  • Support secure over-the-air updates and fleet management
  • Operate without cloud dependency, supporting isolated and air-gapped systems

Use case: Multiple ports have been able to deploy Darktrace directly into the crane’s switch enclosure, securing lateral movement paths without interfering with the crane control software itself.

Segmentation enforcement & real-time threat containment

Darktrace visualizes real-time connectivity and attack pathways across IT, OT, and IoT it and integrates with firewalls (e.g., Fortinet, Cisco, Palo Alto) to enforce segmentation using AI insights alongside Darktrace’s own native autonomous and human confirmed response capabilities.

Benefits of autonomous and human confirmed response:

  • Auto-isolate rogue devices before the threat can escalate
  • Quarantine a suspicious connectivity with confidence operations won’t be halted
  • Autonomously buy time for human responders during off-hours or holidays
  • This ensures segmentation isn't just documented but that in the case of its failure or exploitation responses are performed as a compensating control

No reliance on 3rd parties or external connectivity

Darktrace’s supply chain integrity is a core part of its value to critical infrastructure customers. Unlike solutions that rely on indirect data collection or third-party appliances, Darktrace:

  • Uses in-house engineered sensors and appliances
  • Does not require transmission of data to or from the cloud

This ensures confidence in both your cyber visibility and the security of the tools you deploy.

See examples here of how Darktrace stopped supply chain attacks:

Readiness for USCG and Beyond

With a self-learning system that adapts to each unique port environment, Darktrace helps maritime operators not just comply but build lasting cyber resilience in a high-threat landscape.

Cybersecurity is no longer optional for U.S. ports its operationally and nationally critical. Darktrace delivers the intelligence, automation, and precision needed to meet USCG requirements and protect the digital lifeblood of the modern port.

Continue reading
About the author
Daniel Simonds
Director of Operational Technology
Your data. Our AI.
Elevate your network security with Darktrace AI