On the Importance of Difficulty Calibration in Membership Inference Attacks

Lauren Watson, Chuan Guo, Graham Cormode +1

The vulnerability of machine learning models to membership inference attacks has received much attention in recent years. However, existing attacks mostly remain impractical due to…