Patronus: Safeguarding Text-to-Image Models against White-Box Adversaries
This addresses a critical safety issue for users and developers of text-to-image models by providing robust protection against adversarial manipulation, though it is an incremental improvement on existing safety measures.
The paper tackles the problem of white-box adversaries exploiting text-to-image models to generate unsafe images by introducing Patronus, a defensive framework that protects against such attacks while maintaining safe content generation performance, with results confirming resilience against various fine-tuning attacks.
Text-to-image (T2I) models, though exhibiting remarkable creativity in image generation, can be exploited to produce unsafe images. Existing safety measures, e.g., content moderation or model alignment, fail in the presence of white-box adversaries who know and can adjust model parameters, e.g., by fine-tuning. This paper presents a novel defensive framework, named Patronus, which equips T2I models with holistic protection to defend against white-box adversaries. Specifically, we design an internal moderator that decodes unsafe input features into zero vectors while ensuring the decoding performance of benign input features. Furthermore, we strengthen the model alignment with a carefully designed non-fine-tunable learning mechanism, ensuring the T2I model will not be compromised by malicious fine-tuning. We conduct extensive experiments to validate the intactness of the performance on safe content generation and the effectiveness of rejecting unsafe content generation. Results also confirm the resilience of Patronus against various fine-tuning attacks by white-box adversaries.