Scholarlink

<p>Most existing research on High Efficiency Video Coding (HEVC) video steganography is based on the premise that the receiver can acquire an unaltered bitstream. However, videos uploaded to social media platforms (e.g., YouTube, WeChat, TikTok) are prone to be re-encoded. This process can corrupt or entirely remove the information hidden in the original bitstream. To address this issue, this paper proposes a novel scheme that jointly leverages the frequency domain and the compressed domain. In the frequency domain, an invertible neural network is designed to conceal secret message, where the modifications are directly written into the YUV file. In the compressed domain, data is hidden by modifying Coding Unit (CU) partition during the HEVC encoding process. The impacts of recompression and the available embedding positions are presented through detailed statistics and analysis. Experimental results demonstrate that the proposed algorithm maintains high visual quality and security. Crucially, it ensures reliable and accurate extraction of the secret information even after the video has undergone recompression.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Multi - Domain HEVC video steganography based on invertible neural network

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月25日

查看全文
  • Hemin Yin,
  • Yingnan Zhang,
  • Jun Li,
  • Yanzhe Zhang,
  • Min Shi

摘要

Most existing research on High Efficiency Video Coding (HEVC) video steganography is based on the premise that the receiver can acquire an unaltered bitstream. However, videos uploaded to social media platforms (e.g., YouTube, WeChat, TikTok) are prone to be re-encoded. This process can corrupt or entirely remove the information hidden in the original bitstream. To address this issue, this paper proposes a novel scheme that jointly leverages the frequency domain and the compressed domain. In the frequency domain, an invertible neural network is designed to conceal secret message, where the modifications are directly written into the YUV file. In the compressed domain, data is hidden by modifying Coding Unit (CU) partition during the HEVC encoding process. The impacts of recompression and the available embedding positions are presented through detailed statistics and analysis. Experimental results demonstrate that the proposed algorithm maintains high visual quality and security. Crucially, it ensures reliable and accurate extraction of the secret information even after the video has undergone recompression.