On InstaHide, Phase Retrieval, and Sparse Matrix Factorization
Sitan Chen, Xiaoxiao Li, Zhao Song, Danyang Zhuo
Abstract
In this work, we examine the security of InstaHide, a scheme recently proposed by \cite{hsla20} for preserving the security of private datasets in the context of distributed learning. To generate a synthetic training example to be shared among the distributed learners, InstaHide takes a convex combination of private feature vectors and randomly flips the sign of each entry of the resulting vector with probability 1/2. A salient question is whether this scheme is secure in any provable sense, perhaps under a plausible complexity-theoretic assumption. The answer to this turns out to be quite subtle and closely related to the average-case complexity of a multi-task, missing-data version of the classic problem of phase retrieval that is interesting in its own right. Motivated by this connection, under the standard distributional assumption that the public/private feature vectors are isotropic Gaussian, we design an algorithm that can actually recover a private vector using only the public vectors and a sequence of synthetic vectors generated by InstaHide.
BibTeX
@inproceedings{
chen2021on,
title={On InstaHide, Phase Retrieval, and Sparse Matrix Factorization},
author={Sitan Chen and Xiaoxiao Li and Zhao Song and Danyang Zhuo},
booktitle={International Conference on Learning Representations},
year={2021},
url={https://openreview.net/forum?id=AhElGnhU2BV}
}