Robbing the FED: Directly Obtaining Private Data in Federated Learning with Modified Models
This repo contains a barebones implementation for the attack detailed in the paper:
Fowl L, Geiping J, Czaja W, Goldblum M, Goldstein T.
Robbing the Fed: Directly Obtaining Private Data in Federated Learning with Modified Models.
arXiv preprint arXiv:2110.13057. 2021 Oct 25.
Left: batch of 64 ImageNet images. Right: Images reconstructed with imprint module containing 128 bins placed in front of a ResNet-18. Average PSNR: 70.94.
Abstract:
Federated learning has quickly gained popularity with its promises of increased
user privacy