Hypotheses for the generalization abilities of deepfake detectors.Abstract
This paper aims to explain the generalization of deepfake detectors from the novel perspective of multi-order interactions among visual concepts. Specifically, we propose three hypotheses. 1. Deepfake detectors encode multiorder interactions among visual concepts, in which the low-order interactions usually have substantially negative contributions to deepfake detection. 2. Deepfake detectors with better generalization abilities tend to encode loworder interactions with fewer negative contributions. 3.Generalized deepfake detectors usually weaken the negative contributions of low-order interactions by suppressing their strength. Accordingly, we design several mathematical metrics to evaluate the effect of low-order interaction for deepfake detectors. Extensive comparative experiments are conducted, which verify the soundness of our hypotheses. Based on the analyses, we further propose a generic method, which directly reduces the toxic effects of low-order interactions to improve the generalization of deepfake detectors to some extent.
