Research

Deep Leakage from Gradients

Robustness

Authors

Published on

06/21/2019

Exchanging gradients is a widely used method in modern multi-node machine learning system (e.g., distributed training, collaborative learning). For a long time, people believed that gradients are safe to share: i.e., the training data will not be leaked by gradient exchange. However, we show that it is possible to obtain the private training data from the publicly shared gradients. We name this leakage as Deep Leakage from Gradient and empirically validate the effectiveness on both computer vision and natural language processing tasks. Experimental results show that our attack is much stronger than previous approaches: the recovery is pixel-wise accurate for images and token-wise matching for texts. We want to raise people’s awareness to rethink the gradient’s safety. Finally, we discuss several possible strategies to prevent such deep leakage. The most effective defense method is gradient pruning.

Please cite our work using the BibTeX below.


@article{DBLP:journals/corr/abs-1906-08935,
  author    = {Ligeng Zhu and
               Zhijian Liu and
               Song Han},
  title     = {Deep Leakage from Gradients},
  journal   = {CoRR},
  volume    = {abs/1906.08935},
  year      = {2019},
  url       = {http://arxiv.org/abs/1906.08935},
  archivePrefix = {arXiv},
  eprint    = {1906.08935},
  timestamp = {Mon, 24 Jun 2019 17:28:45 +0200},
  biburl    = {https://dblp.org/rec/journals/corr/abs-1906-08935.bib},
  bibsource = {dblp computer science bibliography, https://dblp.org}
}

Close Modal