Are Watermarked Images Editable? SafeMark for Watermark-Preserving Text-Guided Image Editing

TL;DR

SafeMark is a novel framework that enables watermark-preserving text-guided image editing, ensuring high watermark integrity without compromising semantic edit quality across various editing methods.

Contribution

It introduces a watermark-decoding loss into diffusion-based editors, theoretically grounding watermark preservation and demonstrating its effectiveness empirically.

Findings

SafeMark achieves high watermark bit accuracy across diverse edits.

Semantic quality of images remains high with watermark preservation.

Robustness to post-edit distortions is maintained.

Abstract

This paper investigates a fundamental yet underexplored question: can watermarked images remain editable without compromising watermark integrity? We propose SafeMark, a framework for watermark-preserving text-guided image manipulation that explicitly integrates watermark integrity into the editing process. Specifically, SafeMark adds a thresholded watermark-decoding loss directly to the diffusion editor's training objective, fine-tuning the editor so that semantically valid edits also preserve the embedded watermark at the final output. This design admits a clean information-theoretic justification: maintaining high bit-accuracy on the edited image lower-bounds the mutual information that the editor channel preserves between watermark and edited output, the quantity that fundamentally controls watermark recoverability. SafeMark is compatible with differentiable diffusion-based editors, and requires no architectural modification. Extensive evaluations across multiple datasets, text-guided editing methods, and post-edit distortion settings demonstrate that SafeMark achieves high watermark bit accuracy across diverse editing settings while maintaining high-quality semantic edits, without sacrificing robustness to common post-edit distortions. These results demonstrate that semantic editability and watermark integrity are fundamentally compatible, enabling trustworthy image provenance in generative editing pipelines.