MIRA: Cracking Black-box Watermarking on Deep Neural Networks via Model Inversion-based Removal Attacks

Open in new window