WhatsApp compressed video source camera identification based on photo response nonuniformity

Photo response nonuniformity (PRNU) noise is a unique noise introduced to optical imaging sensors during imaging and can be effectively applied to the source camera identification of compressed video. Due to the problem that existing algorithms do not produce significant effect on extracting PRNU of compressed video, an improved algorithm to extract PRNU was proposed. Firstly, the loop filter of video codec was removed, and the video frame was decomposed by double density-dual tree-complex wavelet transform. Then, the high frequency subband was estimated by bivariate shrinkage algorithm based on Bayesian threshold estimation, and the adaptive window Wiener filter was used for secondary estimation. Finally, after the noise residuals were obtained, they were aggregated by the maximum likelihood estimation method based on quantization parameter weighting, and the PRNU was estimated with video frames. Experiments on the VISION dataset show that the accuracy of the proposed PRNU extraction method in WhatsApp compressed video recognition is improved to 75% at 20 s.