Differential privacy
Also known as: DP · differential privacy
Differential privacy (DP) is a mathematical guarantee: the output barely changes whether or not any single person’s data was included.
- Achieved by adding calibrated noise.
- A provable privacy bound, not a heuristic.
- Used to train models without memorizing individuals.
Why it's special: DP gives a *proof* of privacy rather than a hope. It's the gold standard when you must show, not just claim, that individuals are protected.