Artifact Based Deepfake Detection Methods
Preeti Preeti, Sandhya Rani Bansal · 2023
One of the biggest innovations of Artificial Intelligence (AI) is the ability to generate manipulated or synthesized media (images, videos). When these generated media is created in to look like real and original people then it is called a Deepfake. Deepfakes have gained attention due to their potential to create convincing and misleading content that can be difficult to distinguish from authentic media. While they have garnered positive value in entertainment sector but have seen as a biggest threat regarding its ethical and societal implications particularly in the context of misinformation, identity theft, privacy invasion, and defamation. Rigorous research has been done since its beginning to prevent and detect media forgery and many detection techniques have been developed and discussed in literature. There is no clear winner in identification of DeepFakes however the artefact-based approaches seem to be much superior in the literature. In order to find the lean and identify best detection technique, an evaluation of existing techniques is needed which is the main purpose of our paper. This compares the video forgery detection techniques in terms of convergence, accuracy and equal error rate (EER).