Synthetic Realities: A Critical Review of Deepfake Techniques

Deepfakes involve the manipulation of visual content using various techniques, initially intended for enjoyment, but potentially resulting in detrimental effects for economic, political, and societal objectives, including money fraud and identity fraud. Deepfakes, or hyper-realistic generated pictures and videos, are now feasible due to the fusion of deep learning and computer vision methodologies such as Generative Adversarial Networks (GANs) and autoencoders. Individuals with malicious intent or even non-technical users of machine learning can manipulate a photograph or video by modifying the data, resulting in a new image or video that is undetectable to both humans and machines. Public trust in digital media material has decreased due to deepfakes, since individuals can no longer believe the authenticity of the images they see. Research on deepfake detection is essential to maintain public trust in digital multimedia. This article explores multiple deepfake detection strategies to deal with deepfake and present a literature analysis to give an up to date summery of research activities in the field of deepfake detection. This paper provides a thorough examination of deep-fake phenomena and emphasizes several methods for creating and identifying deepfakes.