RDFMark: Robust Dual-Functional Video Watermarking for Tamper Localization in Social Network Transmissions

Malicious video tampering and unauthorized distribution on social networks pose significant challenges in distinguishing authentic from altered content and raise copyright infringement concerns. Although numerous watermarking schemes have been developed to protect copyright, tamper detection and localization have not been adequately addressed. To this end, we propose RDFMark, a dual-functional solution offering a unified framework for copyright protection and tamper detection. RDFMark introduces a new paradigm in video tamper detection by embedding a robust watermark that survives diverse attacks, enabling the extraction of copyright information and tampering cues. Extensive experiments validate the effectiveness of RDF-Mark against common video tampering scenarios encountered during social network transmissions, such as masking, barrage insertion, and cropping. Furthermore, it demonstrates resilience to geometric distortions, strong compression, re-compression, and composite attacks on social networks.