Set-Membership Inference Attacks using Data Watermarking
This addresses privacy and copyright concerns for data owners by enabling detection of non-consensual data use in generative models, though it is incremental as it builds on existing watermarking techniques.
The authors tackled the problem of detecting unauthorized use of image data in generative models by proposing a set-membership inference attack that uses deep image watermarking to reveal watermarks injected into training data, with empirical results showing it as a principled detection approach.
In this work, we propose a set-membership inference attack for generative models using deep image watermarking techniques. In particular, we demonstrate how conditional sampling from a generative model can reveal the watermark that was injected into parts of the training data. Our empirical results demonstrate that the proposed watermarking technique is a principled approach for detecting the non-consensual use of image data in training generative models.