Ungeneralizable Examples

arXiv:2404.14016v1 Announce Type: new
Abstract: The training of contemporary deep learning models heavily relies on publicly available data, posing a risk of unauthorized access to online data and raising concerns about data privacy. Current approaches to creating unlearnable data involve incorporating small, specially designed noises, but these methods strictly limit data usability, overlooking its potential usage in authorized scenarios. In this paper, we extend the concept of unlearnable data to conditional data learnability and introduce \textbf{U}n\textbf{G}eneralizable \textbf{E}xamples (UGEs). UGEs exhibit …

abstract access arxiv concerns cs.cv cs.lg current data data privacy deep learning examples privacy risk small training type usability usage