Reversible Quantization Index Modulation for Static Deep Neural Network Watermarking

TL;DR

This paper introduces a reversible watermarking scheme for static deep neural networks using quantization index modulation, enabling integrity protection without damaging the model.

Contribution

It proposes a novel RDH-based DNN watermarking method with a one-dimensional QIM approach, addressing integrity and authentication challenges.

Findings

Feasibility demonstrated through training loss and accuracy metrics

Superior adaptability compared to existing watermarking methods

Effective protection of DNN integrity and authentication

Abstract

Static deep neural network (DNN) watermarking techniques typically employ irreversible methods to embed watermarks into the DNN model weights. However, this approach causes permanent damage to the watermarked model and fails to meet the requirements of integrity authentication. Reversible data hiding (RDH) methods offer a potential solution, but existing approaches suffer from weaknesses in terms of usability, capacity, and fidelity, hindering their practical adoption. In this paper, we propose a novel RDH-based static DNN watermarking scheme using quantization index modulation (QIM). Our scheme incorporates a novel approach based on a one-dimensional quantizer for watermark embedding. Furthermore, we design two schemes to address the challenges of integrity protection and legitimate authentication for DNNs. Through simulation results on training loss and classification accuracy, we demonstrate the feasibility and effectiveness of our proposed schemes, highlighting their superior adaptability compared to existing methods.