Deepfake Video
Deepfake videos, synthetically generated videos that convincingly impersonate individuals, are a growing concern due to their potential for misuse in misinformation campaigns and fraudulent activities. Current research focuses on developing robust detection methods, employing various architectures including convolutional neural networks (CNNs), transformers, and generative models like GANs and autoencoders, often incorporating multimodal (audio-visual) analysis to identify inconsistencies between audio and visual components. The field's significance lies in its crucial role in safeguarding digital media integrity and combating the spread of disinformation, driving ongoing efforts to improve detection accuracy and generalization across diverse deepfake generation techniques. This includes exploring both machine-based detection and human perception of deepfakes.
Papers
NPVForensics: Jointing Non-critical Phonemes and Visemes for Deepfake Detection
Yu Chen, Yang Yu, Rongrong Ni, Yao Zhao, Haoliang Li
Unmasking Deepfakes: Masked Autoencoding Spatiotemporal Transformers for Enhanced Video Forgery Detection
Sayantan Das, Mojtaba Kolahdouzi, Levent Özparlak, Will Hickie, Ali Etemad