Big Brother is Watching: Proactive Deepfake Detection via Learnable Hidden Face

2025-04-15

AI Summary

To address the poor generalizability of passive deepfake detection methods and the lack of active defense mechanisms, this paper proposes an active detection framework based on learnable implicit facial watermarking. Our method jointly optimizes watermark embedding and extraction in an end-to-end manner. Key contributions include: (1) introducing the novel “face-in-face” paradigm, where a neutral-face template is learned—rather than manually designed—as a secret watermark; (2) designing a semi-fragile, reversible steganographic network that enables tampering-sensitive embedding–extraction闭环; and (3) incorporating self-mixed reconstruction and simulated channel robust training, coupled with multi-scale feature consistency constraints, to unify detection and anti-counterfeiting capabilities. Extensive experiments on multiple mainstream deepfake benchmarks demonstrate that our approach significantly outperforms existing passive and active methods, achieving high detection accuracy and strong robustness against GAN- and diffusion-based generation as well as diverse post-processing attacks.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Active Learning
• Search and Optimization: Learning to Search

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

As deepfake technologies continue to advance, passive detection methods struggle to generalize with various forgery manipulations and datasets. Proactive defense techniques have been actively studied with the primary aim of preventing deepfake operation effectively working. In this paper, we aim to bridge the gap between passive detection and proactive defense, and seek to solve the detection problem utilizing a proactive methodology. Inspired by several watermarking-based forensic methods, we explore a novel detection framework based on the concept of ``hiding a learnable face within a face''. Specifically, relying on a semi-fragile invertible steganography network, a secret template image is embedded into a host image imperceptibly, acting as an indicator monitoring for any malicious image forgery when being restored by the inverse steganography process. Instead of being manually specified, the secret template is optimized during training to resemble a neutral facial appearance, just like a ``big brother'' hidden in the image to be protected. By incorporating a self-blending mechanism and robustness learning strategy with a simulative transmission channel, a robust detector is built to accurately distinguish if the steganographic image is maliciously tampered or benignly processed. Finally, extensive experiments conducted on multiple datasets demonstrate the superiority of the proposed approach over competing passive and proactive detection methods.

Problem

• Proactive deepfake detection via learnable hidden face
• Bridging gap between passive detection and proactive defense
• Robust detector for distinguishing tampered vs benign images

Innovation

• Uses learnable hidden face for detection
• Employs semi-fragile invertible steganography network
• Optimizes secret template during training