Quantifying and Localizing Usable Information Leakage from Neural Network Gradients

Publication
arXiv preprint arXiv:2105.13929