From Covert Hiding to Visual Editing: Robust Generative Video Steganography

Xueying Mao,Xiaoxiao Hu,Wanli Peng,Zhenliang Gan,Qichao Ying,Zhenxing Qian,Sheng Li,Xinpeng Zhang

2024-01-01

Abstract:Traditional video steganography methods are based on modifying the covert space for embedding, whereas we propose an innovative approach that embeds secret message within semantic feature for steganography during the video editing process. Although existing traditional video steganography methods display a certain level of security and embedding capacity, they lack adequate robustness against common distortions in online social networks (OSNs). In this paper, we introduce an end-to-end robust generative video steganography network (RoGVS), which achieves visual editing by modifying semantic feature of videos to embed secret message. We employ face-swapping scenario to showcase the visual editing effects. We first design a secret message embedding module to adaptively hide secret message into the semantic feature of videos. Extensive experiments display that the proposed RoGVS method applied to facial video datasets demonstrate its superiority over existing video and image steganography techniques in terms of both robustness and capacity.

Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

The problem that this paper attempts to solve is that in video steganography, although traditional methods show a certain level in terms of security and embedding capacity, they lack sufficient robustness to resist common distortions in online social networks (OSNs). Specifically, traditional video steganography methods mainly utilize video data redundancy for information hiding by directly modifying pixel values, coding mappings or adaptive distortion functions. However, these methods are easily eliminated by common post - processing operations (such as JPEG compression, Gaussian blur, etc.), and thus are fragile in the face of various distortions encountered during transmission.

For this reason, the paper proposes a Robust Generative Video Steganography Network (RoGVS) based on Generative Adversarial Networks (GAN). This network embeds secret information by modifying the semantic features of videos during the visual editing process. This method not only improves the security and capacity of steganographic information, but also enhances the resistance to various distortions. By designing a secret message embedding module, RoGVS can adaptively hide secret messages into the semantic features of videos. Experimental results show that the application of RoGVS on the facial video dataset demonstrates its superiority over existing video and image steganography techniques in terms of robustness and capacity.