LandmarkBreaker: A proactive method to obstruct DeepFakes via disrupting facial landmark extraction

Yuezun Li, Pu Sun, Honggang Qi, Siwei Lyu

Abstract: Highlights•We propose a new proactive defense method to obstruct DeepFakes by disrupting facial landmark extraction.•Our method disrupts facial landmark extractors with new masking process and strategies to eliminate unimportant gradients.•We conduct experiments on mainstream facial landmark extractors and study our method under various settings.