A literature review and perspectives in deepfakes: generation, detection, and applications

In the last few years, with the advancement of deep learning methods, especially Generative Adversarial Networks (GANs) and Variational Auto-encoders (VAEs), fabricated content has become more realistic and believable to the naked eye. Deepfake is one such emerging technology that allows the creation of highly realistic, believable synthetic content. On the one hand, Deepfake has paved the way for highly advanced applications in various fields like advertising, creative arts, and film productions. On the other hand, it poses a threat to various Multimedia Information Retrieval Systems (MIPR) such as face recognition and speech recognition systems and has more significant societal implications in spreading misleading information. This paper aims to assist an individual in understanding the deepfake technology (along with its application), current state-of-the-art methods and gives an idea about the future pathway of this technology. In this paper, we have presented a comprehensive literature survey on the application of deepfakes, followed by discussions on state-of-the-art methods for deepfake generation and detection for three media: Image, Video, and Audio. Next, we have extensively discussed the architectural components and dataset used for various methods of deepfakes. Furthermore, we discuss the various limitations and open challenges of deepfakes to identify the research gaps in this field. Finally, discuss the conclusion and future directions to explore the potential of this technology in the coming years.