When Does Data Augmentation Help With Membership Inference Attacks?
Yigitcan Kaya 1 Tudor Dumitras, 1
Abstract et al., 2018; Shokri et al., 2017). In particular, this enables
membership inference attacks (MIAs) to find out whether
Deep learning models often raise privacy con-
a data point was in a model’s training set or not. MIAs of-
cerns as they leak information about t ...


雷达卡




京公网安备 11010802022788号







