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 for integrity authentication. Reversible data hiding (RDH) methods offer a potential solution, but existing approaches suffer from limitations in usability, capacity, and fidelity, hindering their practical adoption. In this paper, we propose a secure static DNN watermarking scheme called Secure Difference Contraction (SDC). Our scheme utilizes a one-dimensional quantizer for watermark embedding and employs dithering to ensure key-dependent security, i.e., the watermark cannot be correctly extracted without the secret key used during embedding. Additionally, we design two schemes to address the challenges of integrity protection and legitimate authentication for DNNs. Simulation results on training loss and classification accuracy demonstrate the feasibility and effectiveness of our proposed methods, highlighting their advantages in capacity and fidelity over existing techniques.

Existing reversible audio watermarking (RAW) techniques are often vulnerable to intentional or even unintentional attacks on the cover object. This paper proposes a robust RAW scheme based on lattices, which is referred to as Meet-in-the-Middle Embedding (MME). In MME, the lattice quantization errors are properly scaled and added back to the quantized host signals such that the receiver can estimate the cover. Scaling factor serves as a key factor to the reversibility of MME, whose feasible range is rigorously justified. Both theoretically and experimentally, we demonstrate the superiority of MME to improved quantization index modulation (IQIM) in terms of signal-to-watermark ratio (SWR) and generalized signal-to-noise ratio (GSNR). Moreover, simulations show that MME also outperforms other state-of-the-arts in SWR, objective difference grade (ODG), and bit error rate (BER).