Learning Self-Consistency for Deepfake Detection
This work provides a strong improvement in deepfake detection performance for the general public and media organizations, particularly in cross-dataset scenarios where existing methods struggle.
This paper addresses deepfake detection by identifying source feature inconsistencies within forged images. Their method improves the average AUC from 96.45% to 98.05% in in-dataset evaluations and from 86.03% to 92.18% in cross-dataset evaluations across seven popular datasets.
We propose a new method to detect deepfake images using the cue of the source feature inconsistency within the forged images. It is based on the hypothesis that images' distinct source features can be preserved and extracted after going through state-of-the-art deepfake generation processes. We introduce a novel representation learning approach, called pair-wise self-consistency learning (PCL), for training ConvNets to extract these source features and detect deepfake images. It is accompanied by a new image synthesis approach, called inconsistency image generator (I2G), to provide richly annotated training data for PCL. Experimental results on seven popular datasets show that our models improve averaged AUC over the state of the art from 96.45% to 98.05% in the in-dataset evaluation and from 86.03% to 92.18% in the cross-dataset evaluation.