An adaptive video watermarking robust to social platform transcoding and attacks

The flourishing progress of social networks and short video platforms has sparked concerns about the copyright protection of rapidly disseminating videos. Video watermarking has emerged as an effective means to embed copyright information into host videos to protect intellectual property. However, existing methods have limitations on countering social platform transcoding and hybrid attacks. To this end, we propose an adaptive video watermarking scheme. In our approach, we propose a novel adaptive high-frequency quantization (AHFQ) strategy based on intra/inter-frame distortion drift and texture joint features, which mitigates the effects of distortion drift and bolsters the resilience against social platform transcoding and hybrid attacks while maintaining good imperceptibility. We also put forward a novel quadruple synchronization verification (QSV) mechanism to tackle the issue of watermark extraction amidst synchronization attacks, specifically asymmetric cropping. Moreover, we devise a loss function to reconstruct the intricate video operation mechanism, facilitating a deeper comprehension of social platform transcoding and significantly fortifying the robustness of our algorithm in the face of social platform transcoding. Extensive experiments demonstrate that the proposed method achieves improved visual quality and exhibits resilience against social platform transcoding, strong recompression, and hybrid attacks, including synchronization attacks (e.g., cropping, frame dropping) accompanied by strong compression.