An efficient motion reference structure based selective encryption algorithm for H.264 videos

In this study, based on both the prediction mechanism of H.264 encoder and the syntax of H.264 bitstream, an efficient selective video encryption algorithm is proposed. The contributions of the study include two aspects. First, motion reference ratio (MRR) of macroblock (MB) is proposed to describe the inter-frame dependency among the adjacent frames. At the MB layer, MRRs of MBs are statistically analysed, and MBs to be encrypted are selected based on the statistical results. Second, at the bitstream layer of MBs, bit-sensitivity is proposed to represent the degree of importance of each bit in the compressed bitstream for reconstructed video quality. The most significant bits for reconstructed video quality are selected to be encrypted based on the bit-sensitivity of H.264 bitstream. The intra-prediction mode codewords, the sign bits of the non-zero coefficients and the info_suffix of motion vector difference codewords are extracted to be encrypted. The proposed two-layer selection scheme improves the encryption efficiency significantly. Experimental results demonstrate that both perceptual security and cryptographic security are achieved, and compared with the existing SEH264 algorithm, the proposed selective encryption algorithm can reduce the computational complexity by 50% on average.