Motion Vector Domain Video Steganography Maintaining the Statistical Characteristics of Skipped Macroblocks
Jun Li, Minqing Zhang, Zhen Zhang · 2023
Motion vector (MV) based video steganography methods achieve covert communication by embedding secret messages in MVs. The most crucial evaluation metric of steganography is the security against steganalysis. However, recently proposed H.264/AVC steganalysis methods based on Skipped macroblocks can effectively capture the statistical differences of the stego video before and after recompression calibration, thus posing challenges to steganography algorithms. In this paper, we propose a distortion function derivation scheme that can maintain the statistical characteristics of the Skipped macroblock of stego videos. Firstly, the designed distortion function upgrades the existing basic distortion function. Secondly, the designed distortion derivation scheme is mainly used to keep the two statistical properties, the motion vector prediction (MVP) and the partition status, unchanged for the Skipped macroblock before and after message embedding. Finally, the derivation method is applied to two typical basic distortion functions for experimental validation. The experimental results show that the proposed derived distortion function can increase the ability to resist the attack of Skipped-based steganalysis.