Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

RingID: Rethinking Tree-Ring Watermarking for Enhanced Multi-Key Identification

About

We revisit Tree-Ring Watermarking, a recent diffusion model watermarking method that demonstrates great robustness to various attacks. We conduct an in-depth study on it and reveal that the distribution shift unintentionally introduced by the watermarking process, apart from watermark pattern matching, contributes to its exceptional robustness. Our investigation further exposes inherent flaws in its original design, particularly in its ability to identify multiple distinct keys, where distribution shift offers no assistance. Based on these findings and analysis, we present RingID for enhanced multi-key identification. It consists of a novel multi-channel heterogeneous watermarking approach designed to seamlessly amalgamate distinctive advantages from diverse watermarks. Coupled with a series of suggested enhancements, RingID exhibits substantial advancements in multi-key identification. Github Page: https://github.com/showlab/RingID

Hai Ci, Pei Yang, Yiren Song, Mike Zheng Shou• 2024

Related benchmarks

TaskDatasetResultRank
Image WatermarkingMS-COCO
PSNR8.66
28
Watermark GenerationCOCO
PSNR12.21
21
Image GenerationStable Diffusion (SD) 1.4
CLIP Score0.3103
18
Image WatermarkingDiffDB
PSNR12.53
17
Watermark DetectionStable Diffusion Original 2-1
AUC1
16
Watermark DetectionStable Diffusion Avg-Removal 2.1
AUC47.19
16
Watermark RobustnessGustavosta Stable-Diffusion-Prompts (test)
Average Robustness88.77
15
Generative Image WatermarkingStable Diffusion Server Scenario v1-5 (test)
CLIP Score0.3637
10
Watermark AttackStable-Diffusion-Prompts
Clean Scenario Performance0.00e+0
9
Watermark DetectionStable Diffusion Avg-Forgery 2-1
AUC100
8
Showing 10 of 25 rows

Other info

Follow for update