A Comprehensive Review of Deepfake Detection In Advanced Neural Network Architectures and Deep Learning Strategies
Deepanshu Singh, Prabhdeep Singh, Rahul Bhandari · 2024
The proliferation of deepfakes, facilitated by advancements in machine learning and artificial intelligence, poses a significant challenge to online security and information integrity. This study reviews current deep fake detection techniques, focusing on methodologies for identifying manipulated content in images, audio, and video. 17 articles were selected for this study using Prisma guidelines. The function of sophisticated machine learning models, such as Long Short-Term Memory (LSTM) networks and (CNNs), is highlighted for automating detection processes. The review also examines notable algorithms and datasets, including FaceForensics++, MesoNet, and XceptionNet, and their effectiveness in enhancing detection accuracy. A case study using frame extraction and resizing techniques demonstrates a detection accuracy of 62%, with precision at 61 %, recall at 63%, and an F1 score of 62% is also included. The findings underscore the need for ongoing refinement and adaptation of detection systems to counteract the evolving nature of deepfake technologies. The study providing thorough overview of the state-of-the-art in deepfake detection, offering insights into effective methodologies and future research directions.