کنفرانس مهندسی برق ایران

صفحه اصلی / سی و سومین کنفرانس بین المللی مهندسی برق

A Robust Video Steganography using 3D-CNN and Maximum Mean Discrepancy Cost Function

نویسندگان :

Ali Ghofrani¹ Rahil Mahdian Toroghi² Hassan Zareian³

1- Iran Broadcasting University 2- Iran Broadcasting University 3- Iran Broadcasting University

کلمات کلیدی :

Video steganography,،3D-CNN,،Maximum Mean Discrepancy (MMD)،Spatio-temporal features

چکیده :

Video steganography aims at embedding a secret video within a cover video so that the output container does not attract any attentions. Furthermore, the quality of the container should resemble the original cover in order to conceal the existence of any hidden information, no matter whether the hidden video is encrypted or not. In this paper, a 3D-CNN has been proposed to ensure that the distribution of the container video frames closely resembles the original cover, confirming the imperceptibility of the embedded information. This is achieved by leveraging a novel optimization function, namely MMD to match the distribution of the container video frames to that of the cover. Similarly, it pursues the same density functions for the reveal and secret paired video, as well. The UCF-101 dataset has been employed for training of the model. The evaluation measures of SSIM, PSNR, APD and VIF have been involved and the outputs are further visually observed. The results clearly indicate that the proposed model outperforms the competitive models using the novel cost function.