Probabilistically Robust Watermarking of Neural Networks
This addresses the problem of verifying ownership of stolen models in MLaaS platforms, offering a robust solution against common attacks, though it is incremental in improving existing watermarking methods.
The paper tackles the vulnerability of deep learning watermarks to model stealing attacks by introducing a trigger set-based watermarking approach that is resilient to functionality stealing, such as extraction and distillation, without requiring additional training and applicable to any model architecture. It shows that the method outperforms current state-of-the-art techniques across multiple benchmarks.
As deep learning (DL) models are widely and effectively used in Machine Learning as a Service (MLaaS) platforms, there is a rapidly growing interest in DL watermarking techniques that can be used to confirm the ownership of a particular model. Unfortunately, these methods usually produce watermarks susceptible to model stealing attacks. In our research, we introduce a novel trigger set-based watermarking approach that demonstrates resilience against functionality stealing attacks, particularly those involving extraction and distillation. Our approach does not require additional model training and can be applied to any model architecture. The key idea of our method is to compute the trigger set, which is transferable between the source model and the set of proxy models with a high probability. In our experimental study, we show that if the probability of the set being transferable is reasonably high, it can be effectively used for ownership verification of the stolen model. We evaluate our method on multiple benchmarks and show that our approach outperforms current state-of-the-art watermarking techniques in all considered experimental setups.