Multimedia Deepfake Detection

In tech-enabled communities, social media allows users to access multimedia content easily. With recent advancements in computer vision and natural language processing, machine learning (ML) and deep learning (DL) models have evolved. With advancements in generative adversarial networks (GAN), it has become possible to create synthetic media of a person or use some person’s contents to fit other environments. Deepfakes are fake media generated using advanced tools and applications, which may mislead people and create an issue of trust within communities. Detecting fakes is crucial and essential for maintaining trust in the digital world. This paper aims to review the current landscape of deepfake generation, detections, methodologies, and implementation based on certain factors. It also gives an idea about the risks, accuracy, and efficiency related to each work. This paper analyzes a variety of methods put out in recent literature to provide a thorough evaluation of the body of research on deepfake detection and also proposes a system. We rigorously divide these strategies into three categories: feature-based, temporal-based, and deep feature-based. We next evaluate each technique’s detection accuracy, computing efficiency, and ability to adjust to changing deepfake methods. Important issues are discussed in our comparison analysis.