High-quality public datasets significantly prompt the prosperity of deep neural networks (DNNs).

High-quality public datasets significantly prompt the prosperity of deep neural networks (DNNs).

High-quality public datasets significantly prompt the prosperity of deep neural networks (DNNs). In this paper, we revisit existing DOV methods and find that they all mainly focused on the first stage by designing different types of dataset watermarks and directly exploiting watermarked samples as the verification samples for ownership verification. As such, their success relies on an underlying assumption that verification is a \emph{one-time} and \emph{privacy-preserving} process, which does not necessarily hold in practice. To alleviate this problem, we propose \emph{ZeroMark} to conduct ownership verification without disclosing dataset-specified watermarks. Our method is inspired by our empirical and theoretical findings of the intrinsic property of DNNs trained on the watermarked dataset.

ZeroMark, a defense startup based in the United States, thinks it has a solution. It wants to turn the rifles of frontline soldiers into "handheld Iron Domes." The idea is simple: Make it easier to shoot a drone out of the sky with a bullet. The problem is that drones are fast and maneuverable, making them hard for even a skilled marksman to hit. ZeroMark's system would add aim assistance to existing rifles, ostensibly helping soldiers put a bullet in just the right place.