DICTION: DynamIC robusT whIte bOx watermarkiNg scheme (2210.15745v1)

Abstract: Deep neural network (DNN) watermarking is a suitable method for protecting the ownership of deep learning (DL) models derived from computationally intensive processes and painstakingly compiled and annotated datasets. It secretly embeds an identifier (watermark) within the model, which can be retrieved by the owner to prove ownership. In this paper, we first provide a unified framework for white box DNN watermarking schemes. It includes current state-of-the art methods outlining their theoretical inter-connections. In second, we introduce DICTION, a new white-box Dynamic Robust watermarking scheme, we derived from this framework. Its main originality stands on a generative adversarial network (GAN) strategy where the watermark extraction function is a DNN trained as a GAN discriminator, and the target model to watermark as a GAN generator taking a GAN latent space as trigger set input. DICTION can be seen as a generalization of DeepSigns which, to the best of knowledge, is the only other Dynamic white-box watermarking scheme from the literature. Experiments conducted on the same model test set as Deepsigns demonstrate that our scheme achieves much better performance. Especially, and contrarily to DeepSigns, with DICTION one can increase the watermark capacity while preserving at best the model accuracy and ensuring simultaneously a strong robustness against a wide range of watermark removal and detection attacks.