Real-time and screen-cam robust screen watermarking

Unauthorized photography of screen-loaded information has become a low-cost and difficult-to-trace method for leakage. Many researchers have proposed screen-cam robust watermarking methods for digital images and documents to identify surreptitious photography. However, these methods are designed for specific types of data and cannot be used in real time, which limits their applicability. This study proposes a screen watermarking method that possesses screen-cam robustness and real-time watermark refreshing. Specifically, a resolution-adaptive watermark template was designed that comprises message watermark templates and two types of synchronization watermark templates. To achieve real-time watermark refreshing, the pregenerated watermark template is directly embedded into the back buffer, which represents the content to be loaded onto the screen. A simplified just noticeable difference model was employed to ensure watermark imperceptibility. For watermark extraction, the positions of the embedded synchronization watermark templates are calculated using a synchronization response index method. Based on these positions, we can perform automatic perspective correction and locate the message watermark template embedded region. Finally, the message watermark was extracted from the discrete Fourier transform domain of the noise component. Extensive experiments demonstrated that the watermark refresh rate meets practical requirements and that the proposed method is robust against image-processing attacks and screen-cam attacks. •We implement a screen-cam robust watermark capable of real-time screen loading.•A watermark template composed of message and synchronization watermark is designed.•We run watermark embedding process within the Desktop Window Manager framework.•The watermark is directly embedded into the back buffer through hooking swap chains.•The method can achieve automatic perspective correction and blind extraction.