Video-Tampering Detection and Content Reconstruction via Self-Embedding

Abstract

Rapidly improving video-editing software tools and algorithms have made video content manipulation and modification feasible even by inexpert users. Detecting video tampering and recovering the original content of the tampered videos is, thus, a major need in many applications. Although detection and localization of the tampering in certain types of video editing have successfully been addressed in the literature, attempts for recovering the tampered videos are bound to methods using watermarks. In this paper, a scheme for the reconstruction of the tampered video through watermarking is proposed. The watermark payload, which consists of highly compressed versions of keyframes of the video and localization information, is embedded in the video using fountain coding. In addition to tampering localization, the proposed scheme can subsequently recover the content of the original video that has undergone malicious attacks. The proposed reconstruction method is quite capable of recovering the video for tampering rate as high as 67% and outperforms the latest tampered video recovery schemes in terms of both video quality and tampering percentage.