Hiding Data Within Thumbnail Videos: An Adaptive Downsampling-Resilient Video Steganography Method

The rapid evolution of social media platforms presents novel opportunities and challenges for video steganography techniques. This paper focuses on taking thumbnail videos as the medium for data hiding, which are extensively utilized across social media platforms. The investigation into the method of embedding secret data into post-downsampled video leads to the proposal of an adaptive, downsampling-resistant method of video steganography, tailored to leverage the unique characteristics of these widely-used video formats. To obfuscate the detectability of the target frames, a pseudorandom number algorithm is employed that adaptively selects the target frames based on a secret key and the characteristics of the video itself. To mitigate the visual perturbations caused by data embedding, the luminance channel of the video serves as the embedding medium, and Syndrome-Trellis Codes (STC) are utilized to compute the embedding pixel positions before and after downsampling separately. Extensive evaluations of this method across various downsampling algorithms and scaling ratios were conducted. The experimental results demonstrate the method’s effectiveness in ensuring that the hidden data remains undetectable and intact.