A High-Performance CNN-Applied HEVC Steganography Based on Diamond-Coded PU Partition Modes

Abstract

High efficiency video coding (HEVC) is the latest high-performance video coding standard, and HEVC video steganography has become a new way to hide data for covert communication. This paper proposes a novel multilevel steganography algorithm based on diamond-encoded prediction unit (PU) partition modes. The PU modes of smaller <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$8\times 8$</tex-math></inline-formula> and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$16\times 16$</tex-math></inline-formula> CUs are selected as carriers for information hiding. The diamond-coding rules are adopted to enhance the expressive ability of limited PU types, allowing them to carry more information under limited modification. Based on the encoded PU partition modes, three different embedding levels with different capacities are proposed. This paper’s most outstanding contribution is the introduction of convolutional neural networks (CNNs) for the first time to improve visual quality and reduce steganographic video bitrate increases. Experimental results show that the embedding capacity of the proposed algorithm is significantly higher than the state-of-the-art work at the same bitrate, whether in high- or low-resolution HEVC videos. At the same time, the visual quality of the steganographic videos is excellent, and the resistance to video steganalysis is strong.