Development of Machine Learning Frame work For Video Forgery Detection

Abstract

Video forgery can be defined as the modification of the video contents. The alteration of newlinethe video by deletion and modification in the sequence of frames is a trivial task, making newlineauthentication and originality detection more important. Frame insertion and deletion are newlinethe most common type of video forgery. Video footage can now be flawlessly changed newlineby utilizing video editing software. As a result, an automatic technique to detect forgeries newlinein video footage is required. Therefore the work is divided into phases that use various newlineapproaches to solve the issues. The work in the thesis begins with the detection of video newlineforgery. Then, progresses towards classifying the forgery present in the videos, like newlinewhether someone inserted or deleted the frames within the video data. This work also newlineidentified the location of the forgery in the video. It is a very formidable job to newlineauthenticate the integrity of the proof; else, manipulated digital media content such as newlinevideos and images can be utilized as a piece of false evidence and induce a complex legal newlineproblem. Hence this work explains some approaches to detecting inter-frame video newlineforgery utilizing the direct mathematical model, machine learning and deep neural newlinenetwork model, and detailed analytical computations. We only use a deep neural network newlinemodel for the deep feature extraction part, further used in forgery detection. This work newlinecalculates the correlation coefficient from the deep features of the adjacent frames rather newlinethan estimating directly from the frames. We divide the procedure of forgery detection newlineinto two phases video forgery detection and video forgery classification. In video newlineforgery detection, we detect input video is original or not. If the video is not authentic, newlinethen the video is checked in the next phase, video forgery classification. In the video newlineforgery classification, we review the forged video for insertion, deletion forgery, and also newlineagain check for originality. The proposed works are generalized, and validated with newlinevarious datasets.