What is DLP and how does it spot data leaks?

The short answer

DLP (data loss prevention) is software that checks what leaves a company, like an email or a USB copy, and warns or blocks when it matches sensitive data such as card numbers.

Animated diagram: What is DLP and how does it spot data leaks?

DLP

DLP (data loss prevention) is software that checks data on its way out of a company. You may have met it at work. In Outlook, a bar appears at the top of a message you’re writing, above the recipients, and says: “Your email conflicts with a policy in your organization.” That bar is a policy tip. The software read your draft and found something that looked sensitive before it left.

Microsoft Purview is one place where companies set this up. DLP doesn’t stop at email. It can also guard enrolled laptops, where people copy files to USB drives or paste text into AI chatbots. The point is simple: sensitive data like card numbers or source code can leave through an ordinary send or paste, and DLP checks those exits.

What DLP actually checks

The company first defines what is sensitive. For a card number, Purview looks for 14 to 19 digits, in a major card brand’s format, that pass the Luhn test. The Luhn test is a math check built on a check digit calculated modulo 10. Reserved test numbers are not flagged.

A matching number alone isn’t strong proof. Because the check is modulo 10, about one random digit string in ten passes it. So DLP looks at context. If a keyword like security code, card number or expiry date, or an expiration date, sits within 300 characters of the number, confidence rises from low to high. In Purview’s definition, the number alone scores 65, and the number plus that context scores 85.

How a DLP policy works

  1. Sensitive data is defined. The company picks what counts as sensitive, such as card numbers, and how to recognize them.
  2. Context raises the confidence. A nearby keyword or expiration date turns a low-confidence match into a high-confidence one.
  3. A rule sets the response. The policy chooses locations, like email or laptops, and the rule holds conditions and actions. Actions include a policy tip, a block you can override with a business justification, or a block with no override.
  4. Every match is recorded. Monitored activities go into the Microsoft 365 Audit log and Activity explorer, where a match shows as DLP rule matched.

The result is a check at the exits: a pattern plus context, compared against a rule, with a record of what happened. On enrolled Windows and macOS laptops, Endpoint DLP adds the same kind of check to copies onto USB removable media and pastes into supported browsers. That’s the kind of exit used in 2023, when a Samsung engineer pasted the source code of a database program into ChatGPT to get a fix.

DLP is not leak-proof

Many people assume DLP catches every leak. It doesn’t. It only sees the channels it’s plugged into. That’s not only company devices: Purview can also inspect web traffic inline through Edge for Business and network integrations. But if someone takes a photo of the screen, no watched channel ever carries the data, so it slips past. DLP matches patterns plus context, but only at the exits it controls.

What to do

  • At home, set Gmail’s Undo Send to 30 seconds: Settings, See all settings, General, Undo Send, then Save changes. The default is only 5 seconds.
  • If a policy tip appears at work, stop and check what you’re sending before you give a business reason to send it anyway.
  • Think twice before pasting source code or customer data into an AI chatbot.
  • Remember that DLP sees only the channels it watches, so it’s not a reason to relax with sensitive data.

#DataLossPrevention#DLP#MicrosoftPurview#DataLeak#ShadowAI#InsiderThreat#DataProtection#InfoSec#Cybersecurity

Watch it in 60 seconds