PersGuard: Preventing Malicious Personalization in Text-to-Image Diffusion Models via Model Backdoors
Summary
Diffusion models (AI systems that generate images from text descriptions) can be misused to create unauthorized portraits or copies of artistic styles through personalization, which threatens privacy and copyright. PersGuard is a new defense framework that embeds protective backdoors (hidden mechanisms) into these models before release, so that if someone tries to personalize the model with protected images, it generates predetermined protective outputs instead, while still working normally for unprotected images.
Solution / Mitigation
PersGuard embeds protective backdoors into pre-trained diffusion models before release. The framework uses three optimization objectives: a backdoor behavior loss to activate protection, a prior preservation loss to maintain normal generation capabilities, and a novel backdoor retention loss designed to ensure the backdoor remains robust when users fine-tune (customize) the model on protected images.
Classification
Related Issues
Original source: http://ieeexplore.ieee.org/document/11523584
First tracked: July 13, 2026 at 08:03 PM
Classified by LLM (prompt v3) · confidence: 85%