On InstaHide, Phase Retrieval, and Sparse Matrix Factorization

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.

Authors

• Sitan Chen
• Xiaoxiao Li
• Zhao Song 0002
• Danyang Zhuo

Keywords

• Distributed learning
• InstaHide
• phase retrieval
• matrix factorization