Seeing is No Longer Believing: How Deepfakes May Shape the Future of Identity Credibility in Media

Deep learning techniques and frameworks for manipulating or generating identities in multimedia, particularly visual data like images and videos, have advanced to a point where even individuals without specialized technical knowledge can create "deepfakes"and use them in real-world situations. This paper examines the potential threats posed by these advancements, offering an overview of manipulation techniques and state-of-the-art detection methods to counteract them. Additionally, it discusses the requirements, trade-offs, and limitations associated with these specific methods, providing a thorough understanding of the current landscape in deepfake technology and its detection, as well as possible trends for the future.