Proactive Deepfake Detection via Training-Free Landmark Perceptual Watermarks

Tianyi Wang, Mengxiao Huang, Harry Cheng, Xiao Zhang, Zhiqi Shen

Abstract: The Deepfake face manipulation technique has garnered significant public attention due to its impacts on both enhancing human experiences and posing security and privacy threats. Despite numerous passive Deepfake detection algorithms that have been attempted to thwart malicious Deepfake attacks, they mostly struggle with the generalizability challenge when confronted with hyper-realistic synthetic facial images contemporarily. To tackle the problem, this paper proposes a proactive Deepfake detection approach by introducing a novel training-free landmark perceptual watermark, LampMark for short. Firstly, we analyze the structure-sensitive characteristics of Deepfake manipulations and devise a secure and confidential transformation pipeline from the structural representations, i.e. facial landmarks, to binary landmark perceptual watermarks. Subsequently, we present an end-to-end watermarking framework that robustly and imperceptibly embeds and extracts watermarks concerning the images to be protected. Relying on promising watermark recovery accuracies, Deepfake detection is accomplished by assessing the consistency between the content-matched landmark perceptual watermark and the robustly recovered watermark of the suspect Deepfake image. Experimental results demonstrate the superior performance of our approach in watermark recovery and Deepfake detection compared to state-of-the-art methods across in-dataset, cross-dataset, and cross-manipulation scenarios.

Primary Subject Area: [Generation] Social Aspects of Generative AI

Secondary Subject Area: [Experience] Multimedia Applications, [Content] Multimodal Fusion

Relevance To Conference: This paper introduces a robust watermarking framework with facial landmark perceptual watermarks to defend against malicious Deepfake image manipulations proactively. We design a watermark construction pipeline to transfer facial landmarks into binary watermarks and robustly embed them into images for protection against Deepfake. Our study contributes innovative insights to the domain of multimedia forensics. Extensive experiments demonstrate superior performance in watermark recovery and Deepfake detection performance based on the proposed idea.

Submission Number: 682