Why outbound email DLP needs reinventing

In 2025, the global average cost of a data breach fell slightly — but remains substantial at USD 4.44 million (IBM Cost of a Data Breach Report 2025). The headline figure hides a painful reality: many of these breaches stem not from sophisticated hacks, but from simple human error: mis-sent emails, accidental forwarding, or replying with the wrong attachment. Because outbound email is a common channel for sensitive data leaving an organization, the risk posed by everyday mistakes is enormous.

In 2025, 53% of data breaches involved customer PII, making it the most commonly compromised asset (IBM Cost of a Data Breach Report 2025). This makes “protection at the moment of send” essential. A single unintended disclosure can trigger compliance violations, regulatory scrutiny, and erosion of customer trust –consequences that are disproportionate to the marginal human errors that cause them.

Traditional DLP has long attempted to mitigate these impacts, but it relies heavily on perfect labelling and rigid pattern-matching. In reality, data loss rarely presents itself as a neat, well-structured pattern waiting to be caught – it looks like everyday communication, just slightly out of context.

How data loss actually happens

Most data loss comes from frustratingly familiar scenarios. A mistyped name in auto-complete sends sensitive data to the wrong “Alex.” A user forwards a document to a personal Gmail account “just this once.” Someone shares an attachment with a new or unknown correspondent without realizing how sensitive it is.

Traditional, content-centric DLP rarely catches these moments. Labels are missing or wrong. Regexes break the moment the data shifts formats. And static rules can’t interpret the context that actually matters – the sender-recipient relationship, the communication history, or whether this behavior is typical for the user.

It’s the everyday mistakes that hurt the most. The classic example: the Friday 5:58 p.m. mis-send, when auto-complete selects Martin, a former contractor, instead of Marta in Finance.

What traditional DLP approaches offer (and where gaps remain)

Most email DLP today follows two patterns, each useful but incomplete.

• Policy- and label-centric DLP works when labels are correct — but content is often unlabeled or mislabeled, and maintaining classification adds friction. Gaps appear exactly where users move fastest
• Rule and signature-based approaches catch known patterns but miss nuance: human error, new workflows, and “unknown unknowns” that don’t match a rule

The takeaway: Protection must combine content + behavior + explainability at send time, without depending on perfect labels.

Your technology primer: The three pillars that make outbound DLP effective

1) Label-free (vs. data classification)

Protects all content, not just what’s labeled. Label-free analysis removes classification overhead and closes gaps from missing or incorrect tags. By evaluating content and context at send time, it also catches misdelivery and other payload-free errors.

• No labeling burden; no regex/rule maintenance
• Works when tags are missing, wrong, or stale
• Detects misdirected sends even when labels look right

2) Behavioral (vs. rules, signatures, threat intelligence)

Understands user behavior, not just static patterns. Behavioral analysis learns what’s normal for each person, surfacing human error and subtle exfiltration that rules can’t. It also incorporates account signals and inbound intel, extending across email and Teams.

• Flags risk without predefined rules or IOCs
• Catches misdelivery, unusual contacts, personal forwards, odd timing/volume
• Blends identity and inbound context across channels

3) Proprietary DSLM (vs. generic LLM)

Optimized for precise, fast, explainable on-send decisions. A DSLM understands email/DLP semantics, avoids generative risks, and stays auditable and privacy-controlled, delivering intelligence reliably without slowing mail flow.

• Low-latency, on-send enforcement
• Non-generative for predictable, explainable outcomes
• Governed model with strong privacy and auditability

The Darktrace approach to DLP

Darktrace / EMAIL – DLP stops misdelivery and sensitive data loss at send time using hold/notify/justify/release actions. It blends behavioral insight with content understanding across 35+ PII categories, protecting both labeled and unlabeled data. Every action is paired with clear explainability: AI narratives show exactly why an email was flagged, supporting analysts and helping end-users learn. Deployment aligns cleanly with existing SOC workflows through mail-flow connectors and optional Microsoft Purview label ingestion, without forcing duplicate policy-building.

Deployment is simple: Microsoft 365 routes outbound mail to Darktrace for real-time, inline decisions without regex or rule-heavy setup.

A buyer’s checklist for DLP solutions

When choosing your DLP solution, you want to be sure that it can deliver precise, explainable protection at the moment it matters – on send – without operational drag.  

To finish, we’ve compiled a handy list of questions you can ask before choosing an outbound DLP solution:

• Can it operate label free when tags are missing or wrong? 
• Does it truly learn per user behavior (no shortcuts)? 
• Is there a domain specific model behind the content understanding (not a generic LLM)? 
• Does it explain decisions to both analysts and end users? 
• Will it integrate with your label program and SOC workflows rather than duplicate them? 

For a deep dive into Darktrace’s DLP solution, check out the full solution brief.

[related-resource]

What is an Adversary-in-the-middle (AiTM) attack?

Adversary-in-the-Middle (AiTM) attacks are a sophisticated technique often paired with phishing campaigns to steal user credentials. Unlike traditional phishing, which multi-factor authentication (MFA) increasingly mitigates, AiTM attacks leverage reverse proxy servers to intercept authentication tokens and session cookies. This allows attackers to bypass MFA entirely and hijack active sessions, stealthily maintaining access without repeated logins.

This blog examines a real-world incident detected during a Darktrace customer trial, highlighting how Darktrace / EMAIL^TM and Darktrace / IDENTITY^TM identified the emerging compromise in a customer’s email and software-as-a-service (SaaS) environment, tracked its progression, and could have intervened at critical moments to contain the threat had Darktrace’s Autonomous Response capability been enabled.

What does an AiTM attack look like?

Inbound phishing email

Attacks typically begin with a phishing email, often originating from the compromised account of a known contact like a vendor or business partner. These emails will often contain malicious links or attachments leading to fake login pages designed to spoof legitimate login platforms, like Microsoft 365, designed to harvest user credentials.

Proxy-based credential theft and session hijacking

When a user clicks on a malicious link, they are redirected through an attacker-controlled proxy that impersonates legitimate services.  This proxy forwards login requests to Microsoft, making the login page appear legitimate. After the user successfully completes MFA, the attacker captures credentials and session tokens, enabling full account takeover without the need for reauthentication.

Follow-on attacks

Once inside, attackers will typically establish persistence through the creation of email rules or registering OAuth applications. From there, they often act on their objectives, exfiltrating sensitive data and launching additional business email compromise (BEC) campaigns. These campaigns can include fraudulent payment requests to external contacts or internal phishing designed to compromise more accounts and enable lateral movement across the organization.

Darktrace’s detection of an AiTM attack

At the end of September 2025, Darktrace detected one such example of an AiTM attack on the network of a customer trialling Darktrace / EMAIL and Darktrace / IDENTITY.

In this instance, the first indicator of compromise observed by Darktrace was the creation of a malicious email rule on one of the customer’s Office 365 accounts, suggesting the account had likely already been compromised before Darktrace was deployed for the trial.

Darktrace / IDENTITY observed the account creating a new email rule with a randomly generated name, likely to hide its presence from the legitimate account owner. The rule marked all inbound emails as read and deleted them, while ignoring any existing mail rules on the account. This rule was likely intended to conceal any replies to malicious emails the attacker had sent from the legitimate account owner and to facilitate further phishing attempts.

Internal and external phishing

Following the creation of the email rule, Darktrace / EMAIL observed a surge of suspicious activity on the user’s account. The account sent emails with subject lines referencing payment information to over 9,000 different external recipients within just one hour. Darktrace also identified that these emails contained a link to an unusual Google Drive endpoint, embedded in the text “download order and invoice”.

As Darktrace / EMAIL flagged the message with the ‘Compromise Indicators’ tag (Figure 2), it would have been held automatically if the customer had enabled default Data Loss Prevention (DLP) Action Flows in their email environment, preventing any external phishing attempts.

Darktrace analysis revealed that, after clicking the malicious link in the email, recipients would be redirected to a convincing landing page that closely mimicked the customer’s legitimate branding, including authentic imagery and logos, where prompted to download with a PDF named “invoice”.

After clicking the “Download” button, users would be prompted to enter their company credentials on a page that was likely a credential-harvesting tool, designed to steal corporate login details and enable further compromise of SaaS and email accounts.

Darktrace’s Response

In this case, Darktrace’s Autonomous Response was not fully enabled across the customer’s email or SaaS environments, allowing the compromise to progress,  as observed by Darktrace here.

Despite this, Darktrace / EMAIL’s successful detection of the malicious Google Drive link in the internal phishing emails prompted it to suggest ‘Lock Link’, as a recommended action for the customer’s security team to manually apply. This action would have automatically placed the malicious link behind a warning or screening page blocking users from visiting it.

Furthermore, if active in the customer’s SaaS environment, Darktrace would likely have been able to mitigate the threat even earlier, at the point of the first unusual activity: the creation of a new email rule. Mitigative actions would have included forcing the user to log out, terminating any active sessions, and disabling the account.

Conclusion

AiTM attacks represent a significant evolution in credential theft techniques, enabling attackers to bypass MFA and hijack active sessions through reverse proxy infrastructure. In the real-world case we explored, Darktrace’s AI-driven detection identified multiple stages of the attack, from anomalous email rule creation to suspicious internal email activity, demonstrating how Autonomous Response could have contained the threat before escalation.

MFA is a critical security measure, but it is no longer a silver bullet. Attackers are increasingly targeting session tokens rather than passwords, exploiting trusted SaaS environments and internal communications to remain undetected. Behavioral AI provides a vital layer of defense by spotting subtle anomalies that traditional tools often miss

Security teams must move beyond static defenses and embrace adaptive, AI-driven solutions that can detect and respond in real time. Regularly review SaaS configurations, enforce conditional access policies, and deploy technologies that understand “normal” behavior to stop attackers before they succeed.

Credit to David Ison (Cyber Analyst), Bertille Pierron (Solutions Engineer), Ryan Traill (Analyst Content Lead)

Appendices

Models

SaaS / Anomalous New Email Rule

Tactic – Technique – Sub-Technique  

Phishing - T1566

Adversary-in-the-Middle - T1557