Scanning your connection...
Back to Glossary
Data Protection

What is Differential Privacy?

A mathematical framework for sharing aggregate information about a dataset while provably protecting the privacy of individual entries.

Differential privacy adds carefully calibrated noise to data queries, allowing useful statistical analysis while making it impossible to identify any individual's contribution.

How It Works

  • Random noise is added to query results
  • The noise is large enough to hide any individual's data
  • But small enough that aggregate statistics remain accurate
  • Provides a mathematical guarantee of privacy (epsilon parameter)

Real-World Uses

  • Apple: Uses differential privacy in iOS to collect usage statistics
  • Google: RAPPOR system for Chrome usage data
  • US Census: Applied differential privacy to 2020 Census data

The Epsilon Problem

  • Epsilon (ε) measures the privacy guarantee — smaller is more private
  • There's no universal agreement on what epsilon value is "private enough"
  • Apple uses ε = 4-8, while academic researchers often recommend ε < 1
  • Companies may claim differential privacy while using loose parameters

Related Terms

Have more questions?

Use our guided flow to get the right next privacy step for Differential Privacy.

Open Guided Flow