DAWN: Dynamic Adversarial Watermarking of Neural Networks

TL;DR

DAWN introduces a novel dynamic watermarking method that deters model extraction theft by embedding watermarks through API responses, effectively identifying surrogate models with high confidence and minimal accuracy loss.

Contribution

It is the first approach to use dynamic watermarking at the API level to prevent model extraction IP theft without altering training procedures.

Findings

Resilient against state-of-the-art model extraction attacks

Watermarks embedded in surrogate models with high confidence

Negligible impact on model prediction accuracy

Abstract

Training machine learning (ML) models is expensive in terms of computational power, amounts of labeled data and human expertise. Thus, ML models constitute intellectual property (IP) and business value for their owners. Embedding digital watermarks during model training allows a model owner to later identify their models in case of theft or misuse. However, model functionality can also be stolen via model extraction, where an adversary trains a surrogate model using results returned from a prediction API of the original model. Recent work has shown that model extraction is a realistic threat. Existing watermarking schemes are ineffective against IP theft via model extraction since it is the adversary who trains the surrogate model. In this paper, we introduce DAWN (Dynamic Adversarial Watermarking of Neural Networks), the first approach to use watermarking to deter model extraction IP theft. Unlike prior watermarking schemes, DAWN does not impose changes to the training process but it operates at the prediction API of the protected model, by dynamically changing the responses for a small subset of queries (e.g., <0.5%) from API clients. This set is a watermark that will be embedded in case a client uses its queries to train a surrogate model. We show that DAWN is resilient against two state-of-the-art model extraction attacks, effectively watermarking all extracted surrogate models, allowing model owners to reliably demonstrate ownership (with confidence $>1- 2^{-64}$), incurring negligible loss of prediction accuracy (0.03-0.5%).