Deepfake Dilemma: A Review of GAN-Based Detection for Facial Manipulations

Deepfakes, or the synthetic media generated by artificial intelligence, pose a serious threat to the authenticity of the information. The fake visual data available on social media poses a significant threat to the users associated with it. In this review, we look into the usage of GANs for the detection of generated content known as deepfakes. Further, we will understand the mathematical functioning of GANs along with the different architectures like cGANs, DCGANs, StarGANs, and StyleGANs. The detection of these deepfakes has been around for a while, so some detection methods are available; we will explore those as well. Also we will look into the deepfake datasets, which are updated over time with growing quantity and quality of samples. To check how well a method works in detecting the deepfake, we have some evaluation metrics, broadly categorised as quantitative and qualitative measures. This review provides a complete overview of the GAN technologies used for deepfake detection of the visual content.