Separable Reversible Data Hiding in Encrypted 3D Mesh Models Based on Spatial Clustering and Multi-MSB Prediction

Abstract. Reversible data hiding in the encrypted domain (RDH-ED) involves performing data encryption to protect privacy and hiding data for covert communication or access control. Current research mainly focuses on exploring spatial correlation of adjacent vertices and local spatial correlation to reserve embedding space, while ignoring global spatial correlation, limiting embedding capacity. To address this, a method combining spatial clustering with multi-MSB (multiple most significant bit) prediction is proposed to enhance embedding rate and capacity. Input model vertices are partitioned into exclusive clusters through spatial clustering, ensuring close proximity within each cluster. An optimal reference vertex search algorithm is then used to identify the best reference vertex in each cluster to maximize embeddable capacity, and multi-MSB prediction determines the embedding length for each embeddable vertex. The basic embedding length of the reference vertex also serves as additional embedding space, further boosting capacity. Finally, data are embedded in the reserved space through bit substitution. Experimental results demonstrate that the proposed method achieves higher embedding capacity and low computational overhead while supporting reversibility and separability. The proposed scheme provides a robust, high-capacity, and scalable solution for secure data embedding in encrypted 3D graphics, which is increasingly relevant in real-world 3D, geological model, and virtual reality applications.