Yihao LIU

A practical artwork protection method against style mimicry attack, robust to Noise Upscaling

We propose the first label-only membership inference attacks against pre-trained large vision-language models.