Forged Video Detection Using Deep Learning: A SLR

Abstract

In today’s digital landscape, video and image data have emerged as pivotal and widely adopted means of communication. They serve not only as a ubiquitous mode of conveying information but also as indispensable evidential and substantiating elements across diverse domains, encompassing law enforcement, forensic investigations, media, and numerous others. This study employs a systematic literature review (SLR) methodology to meticulously investigate the existing body of knowledge. An exhaustive review and analysis of precisely 90 primary research studies were conducted, unveiling a range of research methodologies instrumental in detecting forged videos. The study’s findings shed light on several research methodologies integral to the detection of forged videos, including deep neural networks, convolutional neural networks, Deepfake analysis, watermarking networks, and clustering, amongst others. This array of techniques highlights the field and emphasizes the need to combat the evolving challenges posed by forged video content. The study shows that videos are susceptible to an array of manipulations, with key issues including frame insertion, deletion, and duplication due to their dynamic nature. The main limitations of the domain are copy-move forgery, object-based forgery, and frame-based forgery. This study serves as a comprehensive repository of the latest advancements and techniques, structured, and summarized to benefit researchers and practitioners in the field. It elucidates the complex challenges inherent to video forensics.

Contour-Assistance-Based Video Matting Localization

“…Existing video forensics methods mainly include: (1) Frame-level analysis [33,39]: detecting possible editing or tampering by analyzing each frame of the video. (2) Metadata analysis [2,23]: metadata contained in the video file, such as shooting time and equipment information, can be used to verify the authenticity of the video. (3) Digital signatures and watermarks [44]: Digital signatures or watermarks are embedded in the video to ensure video integrity.…”