Invisible hints such as logos embedded in images are re-rendered by diffusion models during text-guided editing, enabling phishing and model-poisoning attacks with average success rates of 44.4% and 32.2%.
arXiv preprint arXiv:2503.09669 (2025)
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
fields
cs.CR 2years
2026 2verdicts
UNVERDICTED 2representative citing papers
Adversaries can degrade synthetic data quality via small manipulations such as label flipping or feature-importance interventions, substantially harming downstream model performance and increasing statistical divergence from real data.
citing papers explorer
-
Generate "Normal", Edit Poisoned: Branding Injection via Hint Embedding in Image Editing
Invisible hints such as logos embedded in images are re-rendered by diffusion models during text-guided editing, enabling phishing and model-poisoning attacks with average success rates of 44.4% and 32.2%.
-
Quality Degradation Attack in Synthetic Data
Adversaries can degrade synthetic data quality via small manipulations such as label flipping or feature-importance interventions, substantially harming downstream model performance and increasing statistical divergence from real data.