Detecting Deepfake by Creating Spatio-Temporal Regularity Disruption

TL;DR

This paper introduces a novel deepfake detection method that leverages the disruption of statistical regularities in videos, using pseudo-fake generation and a spatio-temporal enhancement to improve generalization without relying on real fake videos.

Contribution

It proposes a new approach that detects deepfakes by identifying regularity disruptions, enhancing generalization by creating pseudo-fake videos and a spatio-temporal learning block.

Findings

Achieves high detection accuracy on multiple datasets.

Does not require real fake videos for training.

Improves generalization to unseen forgery types.

Abstract

Despite encouraging progress in deepfake detection, generalization to unseen forgery types remains a significant challenge due to the limited forgery clues explored during training. In contrast, we notice a common phenomenon in deepfake: fake video creation inevitably disrupts the statistical regularity in original videos. Inspired by this observation, we propose to boost the generalization of deepfake detection by distinguishing the "regularity disruption" that does not appear in real videos. Specifically, by carefully examining the spatial and temporal properties, we propose to disrupt a real video through a Pseudo-fake Generator and create a wide range of pseudo-fake videos for training. Such practice allows us to achieve deepfake detection without using fake videos and improves the generalization ability in a simple and efficient manner. To jointly capture the spatial and temporal disruptions, we propose a Spatio-Temporal Enhancement block to learn the regularity disruption across space and time on our self-created videos. Through comprehensive experiments, our method exhibits excellent performance on several datasets.