High-Capacity Video-in-Video Steganography using Residual Dense Blocks and CBAM Attention
Abstract
Video steganography hides secret content inside video, but hiding a large payload while keeping the cover clean and the data secure remains difficult. This paper presents an encoder-decoder model that embeds a full RGB secret frame into a cover frame at 24 bits per pixel, frame by frame. The encoder joins the cover and secret into a six-channel input and learns a small additive residual, with Residual Dense Blocks and the Convolutional Block Attention Module preserving fine detail. On Vimeo-90K it reaches a stego PSNR of 45.68 dB and a recovery PSNR of 39.25 dB at a 0.78% bit error rate, results that hold on UCF-101, DAVIS 2017 and recorded video, with embedding at 24.8 frames per second on a single GPU. The stego video changes from frame to frame almost exactly as the cover does, and recovery is unaffected by lossless H.264 and H.265 coding. The stego frames pass six of seven classical statistical tests and are only partly flagged by StegExpose, while a robustness study finds good resistance to mild noise and blur but clear weakness against JPEG compression. The results indicate that residual dense learning with attention is an effective approach to full-frame video hiding.