Abstract:
In recent years, the rapid technological development and the emergence of mobile devices, cameras, etc., in addition to the availability of video production, editing, and formatting programs, made it easy to edit, manipulate, and fake or tamper video. As they know that pictures or videos give more information than texts; Video is a very important medium for transferring information from one place to another. One of the important types of evidence in road accidents and theft crimes. Moreover, when forensic analysis is essential for any video, the availability of origin video may be rare therefore the forensic experts must establish decisions based on the present video (under surveillance) and decide if this video is fake (tampered) or not fake. There are multiple methods to tamper video, including active and blind passive methods. In this research, we tried to combine the behavior of active methods in the process of embedding the halftone current frame of video in the DCT Coefficients of next frame of the same video with the behavior of passive methods by comparing the information embedded after extracting with the information of the current frame to determine whether there is a fake in the video or not and which frame contains tamper. The experimental results of the submitted method showed a huge level of success in locating frames in which falsification or tampering occurred through copying, deletion or insertion, or even if copy-move regions. Also, in proposed method we attempted to post-processing the fake frames using the information included in the subsequent frame, if it is not faked. Finally, the original video, the embedded halftone video, and the tamper (fake) video after post processing were compared using PSNR and SSIM similarity scales. At last, the accuracy and precision scores of tampered and non tampered frames are computed.
