🤖 AI Summary
Existing watermarking methods for image authenticity verification lack robustness against both malicious and benign edits. Method: This paper proposes a semantic-driven active watermarking framework: (1) it employs high-dimensional semantic vectors—generated from image captions—as watermark payloads, embedding them into images to establish a semantic communication channel; (2) it introduces a local confidence-based message recovery evaluation mechanism to bridge the gap between active watermarking and passive forensic analysis; and (3) it jointly optimizes an enhanced HiDDeN-based image encoder-decoder, semantic embedding/extraction networks, and a local confidence modeling module. Results: Experiments demonstrate significant improvements in watermark recovery rates under JPEG compression, cropping, filtering, and adversarial perturbations. Crucially, recovered watermark fidelity exhibits strong correlation with local confidence scores (Pearson’s r > 0.92). The implementation is publicly available.
📝 Abstract
This paper proposes a novel approach towards image authentication and tampering detection by using watermarking as a communication channel for semantic information. We modify the HiDDeN deep-learning watermarking architecture to embed and extract high-dimensional real vectors representing image captions. Our method improves significantly robustness on both malign and benign edits. We also introduce a local confidence metric correlated with Message Recovery Rate, enhancing the method’s practical applicability. This approach bridges the gap between traditional watermarking and passive forensic methods, offering a robust solution for image integrity verification. The code is available at https://github.com/gautierevn/swift_watermarking.