MixCon: Adjusting the Separability of Data Representations for Harder Data Recovery

Xiaoxiao Li, YANGSIBO HUANG, Binghui Peng, Zhao Song, Kai Li

Keywords: Data Recovery, Data Separability, Distributed Deep Learning

Abstract: To address the issue that deep neural networks (DNNs) are vulnerable to model inversion attacks, we design an objective function to adjust the separability of the hidden data representations as a way to control the trade-off between data utility and vulnerability to inversion attacks. Our method is motivated by the theoretical insights of data separability in neural networking training and results on the hardness of model inversion. Empirically, we show that there exist sweet-spots by adjusting the separability of data representation, such that it is difficult to recover data during inference while maintaining data utility.

One-sentence Summary: We investigate the trade-off between data utility and risk of data recovery from the angle of adjusting data separability.

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics

Reviewed Version (pdf): https://openreview.net/references/pdf?id=s3G6EvDDiM