What is Smart Classification (SC)?

Machine learning approach to data leak prevention

With Zivver Smart Classification, you improve the deployment of your email business rules. These rules indicate whether it is necessary to treat the email in question as sensitive.

Legacy deployments only use lists of words and patterns to flag sensitive information like medical, financial, or legal. In practice, however, this method yields false positives and decreases user acceptance.

Smart Classification is a step forward. It uses machine learning to detect the contextual signals that make information sensitive. Our research found that Smart Classification is 3 to 4 times more precise than the legacy methods.

You can use Smart Classification with all Zivver products that support business rules.

How Smart Classification scores a message

Smart Classification scores the full content and context of a message. It never judges a single word or number in isolation. A routine word does not trigger a rule on its own, and a sensitive-looking term does not always trigger one.

Disclaimers are excluded from Smart Classification by default. Zivver recognizes the disclaimer automatically. For more information, see Exclude disclaimer from Smart Classification.

Is a message secured or blocked unexpectedly, or not at all? See Why did Zivver flag or not flag my message?

Start with Smart Classification

Your organization will use Smart Classification by default. No action is required from you.

Users might see that Zivver warns less often but more precisely. Smart Classification might not mark information as sensitive that previously got a sensitive flag.

If you have reasons to use the legacy deployment of the business rules, use the Opt-out form to cancel. If you want to upgrade after you canceled, you can speak to your customer success manager.