The article discusses the development of a new method for detecting deepfakes, known as FACTOR. Traditional methods struggle with detecting new, unseen types of deepfake attacks. FACTOR, however, uses fact checking to verify the authenticity of media, specifically checking the claimed facts (e.g., identity) against the observed media. This method is training-free, easy to implement, and performs better than current techniques in detecting deepfakes. The article emphasizes the need for effective deepfake detection methods due to the rapid advancement of deepfake generation techniques and their potential harm to society.
Publication date: 2 Nov 2023
Project Page: https://github.com/talreiss/FACTOR
Paper: https://arxiv.org/pdf/2311.01458
