Membership inference
Working out whether a specific person’s data was used to train a model.
By studying how confidently a model responds to a record, an attacker may infer that it was in the training set, which can itself reveal sensitive information, such as being a patient in a disease dataset.
On the exam
Differential privacy and limiting output detail reduce the risk.
Related terms
- Differential privacyAdding calibrated noise so results don’t reveal any one person’s data.
- Model extraction and model inversionStealing a model, or recovering its training data, through its outputs.
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.
