Reconstructing Training Data from Model Gradient, Provably

Zihan Wang, Jason D. Lee, Qi Lei

Research output: Contribution to journalConference articlepeer-review

3 Scopus citations

Abstract

Understanding when and how much a model gradient leaks information about the training sample is an important question in privacy. In this paper, we present a surprising result: even without training or memorizing the data, we can fully reconstruct the training samples from a single gradient query at a randomly chosen parameter value. We prove the identifiability of the training data under mild conditions: with shallow or deep neural networks and a wide range of activation functions. We also present a statistically and computationally efficient algorithm based on tensor decomposition to reconstruct the training data. As a provable attack that reveals sensitive training data, our findings suggest potential severe threats to privacy, especially in federated learning.

Original languageEnglish (US)
Pages (from-to)6595-6612
Number of pages18
JournalProceedings of Machine Learning Research
Volume206
StatePublished - 2023
Event26th International Conference on Artificial Intelligence and Statistics, AISTATS 2023 - Valencia, Spain
Duration: Apr 25 2023Apr 27 2023

All Science Journal Classification (ASJC) codes

  • Artificial Intelligence
  • Software
  • Control and Systems Engineering
  • Statistics and Probability

Fingerprint

Dive into the research topics of 'Reconstructing Training Data from Model Gradient, Provably'. Together they form a unique fingerprint.

Cite this