SEAL: Semantic Aware Image Watermarking

2025-03-15

AI Summary

Existing watermarking methods for verifying the authenticity of diffusion-generated images suffer from either distribution distortion or reliance on large pre-stored key databases. Method: We propose a semantic-aware watermarking framework featuring (i) a novel semantics-driven embedding mechanism that implicitly encodes CLIP-derived image semantics into the watermark; (ii) locality-sensitive hashing (LSH) for key self-inference, eliminating the need for a pre-stored key database; and (iii) a semantics-conditioned detection network enabling content-adaptive verification. Results: Our method achieves >98.7% detection accuracy under challenging attacks—including noise reuse, malicious object injection, and semantic manipulation—while maintaining exceptional fidelity (LPIPS < 0.01). It significantly enhances robustness against removal and forgery, and is the first to realize high-fidelity, key-database-free, semantics-aware watermarking for generative images.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Adversarial Learning & Robustness
• Natural Language Processing: Generation

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Web Mining and Content Analysis: Web data provenance, reliability, and authenticity
• Security and Privacy: Data transparency and provenance

Abstract

Generative models have rapidly evolved to generate realistic outputs. However, their synthetic outputs increasingly challenge the clear distinction between natural and AI-generated content, necessitating robust watermarking techniques. Watermarks are typically expected to preserve the integrity of the target image, withstand removal attempts, and prevent unauthorized replication onto unrelated images. To address this need, recent methods embed persistent watermarks into images produced by diffusion models using the initial noise. Yet, to do so, they either distort the distribution of generated images or rely on searching through a long dictionary of used keys for detection. In this paper, we propose a novel watermarking method that embeds semantic information about the generated image directly into the watermark, enabling a distortion-free watermark that can be verified without requiring a database of key patterns. Instead, the key pattern can be inferred from the semantic embedding of the image using locality-sensitive hashing. Furthermore, conditioning the watermark detection on the original image content improves robustness against forgery attacks. To demonstrate that, we consider two largely overlooked attack strategies: (i) an attacker extracting the initial noise and generating a novel image with the same pattern; (ii) an attacker inserting an unrelated (potentially harmful) object into a watermarked image, possibly while preserving the watermark. We empirically validate our method's increased robustness to these attacks. Taken together, our results suggest that content-aware watermarks can mitigate risks arising from image-generative models.

Problem

• Develops a distortion-free watermarking method for AI-generated images.
• Enhances robustness against forgery and unauthorized replication attacks.
• Uses semantic embedding for watermark verification without key databases.

Innovation

• Embeds semantic info directly into watermark
• Uses locality-sensitive hashing for key inference
• Conditions detection on original image content