Toward the Creation and Obstruction of DeepFakes

Yuezun Li, Pu Sun, Honggang Qi, Siwei Lyu

Abstract: AI-synthesized face-swappingDeepFake Face swapping videos, commonly known as DeepFakesDeepFake, is an emerging problem threatening the trustworthiness of online information. The need to develop and evaluate DeepFake detectionDeepFake detection algorithms calls for large-scale datasets. However, current DeepFakeDeepFake datasets suffer from low visual quality and do not resemble DeepFakeDeepFake videos circulated on the Internet. We present a new large-scale challenging DeepFakeDeepFake video dataset, Celeb-DFCeleb-DF, which contains 5, 639 high-quality DeepFakeDeepFake videos of celebrities generated using an improved synthesisSynthesis process. We conduct a comprehensive evaluation of DeepFake detectionDeepFake detection methods and datasets to demonstrate the escalated level of challenges posed by Celeb-DFCeleb-DF. Then we introduce Landmark Breaker, the first dedicated method to disrupt facial landmarkFacial landmark extraction, and apply it to the obstruction of the generation of DeepFakeDeepFake videos. The experiments are conducted on three state-of-the-art facial landmarkFacial landmark extractors using our Celeb-DFCeleb-DF dataset.