Geometry-aware fragile watermarking with cryptographic functions for authenticity verification of glTF 3D models
Artykuł w czasopiśmie
MNiSW
100
Lista 2024
| Status: | |
| Autorzy: | Matczuk Marcin, Traczyński Mateusz |
| Dyscypliny: | |
| Aby zobaczyć szczegóły należy się zalogować. | |
| Rok wydania: | 2026 |
| Wersja dokumentu: | Elektroniczna |
| Język: | angielski |
| Numer czasopisma: | 7 |
| Wolumen/Tom: | 20 |
| Strony: | 139 - 156 |
| Impact Factor: | 1,3 |
| Bazy: | BazTech |
| Efekt badań statutowych | NIE |
| Finansowanie: | FD-20/IT-3/999 |
| Materiał konferencyjny: | NIE |
| Publikacja OA: | TAK |
| Licencja: | |
| Sposób udostępnienia: | Otwarte czasopismo |
| Wersja tekstu: | Ostateczna wersja opublikowana |
| Czas opublikowania: | W momencie opublikowania |
| Data opublikowania w OA: | 1 czerwca 2026 |
| Abstrakty: | angielski |
| Modern graphics processing units possess the requisite computing power to efficiently render complex three-dimensional (3D) scenes. The Graphics Library Transmission Format (glTF) is a file format and a standard that facilitates the efficient transfer of entire virtual worlds between various applications. Despite the growing adoption of the glTF in the context of rendering and transferring 3D content, the area of watermarking within this format remains poorly researched. The article presents a novel scheme for verifying the authenticity of 3D assets in glTF standard. The method is based on a fragile watermarking mechanism and allows for the localisation of sabotage spots. The work employs a Hash-based Message Authentication Code (HMAC) and a BLAKE2 cryptographic hash function to generate components of the watermark. An authorial algorithm called the High-Density Mesh Locator (HDML) has been developed. The HDML identifies areas with dense, complex geometry and leverages a kd-tree data structure to expedite and optimise spatial queries. The selection of vertices located in regions of high geometric complexity effectively obscures the modifications made. The mechanism for detecting sabotage involves segmenting the model using k-means++ algorithm and applying a separate part of the watermark to each segment. The methodology consists of four sequential stages: (1) watermarking various 3D models; (2) making modifications or omitting them; (3) verifying the authenticity of data based on a watermark; and (4) detecting places of interference in the model. The proposed method introduces a minimal level of interference into the geometric structure of the object. This assertion is supported by the obtained Peak Signal-to-Noise Ratio metric of ~210 dB and Hausdorff distance of ~10^(-8). The amalgamation of short hashes and the authorial HDML algorithm yields a highly transparent watermark that allows the detection of a single bit change |
