91影视

Recently, the research team from the Faculty of Engineering of 91影视 (SMBU), in collaboration with Russian research institutions, successfully published a paper at the IEEE International Workshop on Information Forensics and Security (IEEE WIFS 2025), where it was awarded the "Best Paper Award." IEEE WIFS is one of the flagship academic conferences under the IEEE Signal Processing Society (IEEE SPS), focusing on cutting-edge topics such as information forensics, security, and trustworthy artificial intelligence. The conference, organized by IEEE SPS, is known for its strict academic review process and significant international research influence.

The paper was co-authored by Professor Chen Changsheng (corresponding author, second from the right in the picture) and Professor Tan Shunquan from SMBU’s Faculty of Engineering, alongside Master's students from Shenzhen University. It was a joint effort with researchers Yulia Chernyshova, Dmitry Nikolaev, and Vladimir Arlazarov from Smart Engine, a Russian intelligent engineering service company. Dmitry Nikolaev and Vladimir Arlazarov are also affiliated with the Russian Academy of Sciences' Computer Science and Control Research Center (FRC CSC RAS). This research integrates the strengths of both teams: the Chinese team provided advanced AI solutions to address real-world problems, while the Russian collaborators contributed the necessary data support for the study.

The paper addresses the real-world forensic challenges posed by digital documents in screen recapture scenarios, and it innovatively proposes a dual-branch forensic framework combining chromatic features and frequency-domain moiré pattern modeling. By using Masked Attention to suppress complex textures from the document and applying the Frequency-domain Moiré-Aware Adapter to enhance the moiré and other frequency-domain features generated during recapture, the model achieves efficient detection of recapture artifacts on displayed documents. Experimental results show that this method significantly outperforms existing mainstream detection techniques in multiple cross-device and real-world scenario tests, highlighting the model's ability to distinguish between genuine textures and recapture artifacts in complex backgrounds. This breakthrough provides a crucial technological advancement in building high-trust document authentication systems.

This achievement not only showcases SMBU’s research strength in information security and document image forensics but also marks a new level in collaboration between the university and Russian research institutions in these fields. Experimental results show that this method significantly outperforms existing mainstream detection techniques in multiple cross-device and real-world scenario tests, highlighting the model's ability to distinguish between genuine textures and recapture artifacts in complex backgrounds. This breakthrough provides a crucial technological advancement in building high-trust document authentication systems.

Paper Information: P. Li, C. Chen, Y. Chernyshova, D. Nikolaev, S. Tan, and V. Arlazarov, “Disentangling Moiré and Texture: Towards Robust Display-Recapture Detection for Document Images,” 17th International Workshop on Information Forensics and Security (WIFS), Perth, Australia, 2025.