Reversible Data Hiding for 3D Mesh Models in Encrypted Domain Based on Adaptive MSB and Difference Prediction

Mengyuan Zhang, Zihao Fu, Xiaochun Wu, Wanli Lyu

Abstract: Reversible data hiding in encrypted domains (RDH-ED) enables secret data embedding within an encrypted carrier, providing dual protection for both the carrier and the hidden message during transmission. However, most existing research focuses on image-based applications, while current RDH algorithms for 3D models still exhibit limited capacity, especially in vacating room after encryption (VRAE) schemes. We propose a method using adaptive MSB and difference prediction for 3D mesh models. First, the vertex information is preprocessed and encrypted using chunked modulus and the 3D Arnold Transform. Then, the first vertex within each subblock is chosen as the reference vertex, while MSB prediction and difference prediction are employed to predict the remaining vertices within the subblock, generating capacity for secret message embedding. Finally, the separable operations of original model restoration and secret message extraction can be performed by the receiver based on their available keys. The proposed algorithm outperforms all VRAE-based methods in embedding rate while maintaining model security. On the Princeton dataset, the average embedding rate reaches 46.28 bits per vertex.

External IDs:dblp:conf/icic/ZhangFWL25