Differential privacy
Adding calibrated noise so results don’t reveal any one person’s data.
It gives a mathematical guarantee that an output barely changes whether or not one individual’s record is included, which protects against membership inference. There is a trade-off between privacy and accuracy.
On the exam
Differential privacy defends against attacks that infer whether a person was in the training data.
Related terms
- Membership inferenceWorking out whether a specific person’s data was used to train a model.
- Anonymisation and pseudonymisationRemoving or replacing identifiers in personal data.
AAIA Test yourself on this with timed AAIA practice questions in the AAIA Exam Simulation Hub, or read the free AAIA exam guide.
Test yourself with today’s question of the day.
