MMA-Diffusion: MultiModal Attack on Diffusion Models
Authors:
Top Cited Papers
- 16 June 2024
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 7737-7746
Abstract
In recent years, Text-to-Image (T2I) models have seen remarkable advancements, gaining widespread adoption. However, this progress has inadvertently opened avenues for potential misuse, particularly in generating inappropriate or Not-Safe-For-Work (NSFW) content. Our work introduces MMA-Diffusion, a framework that presents a significant and realistic threat to the security of T2I models by effectively circumventing current defensive measures in both open-source models and commercial online services. Unlike previous approaches, MMA-Diffusion leverages both textual and visual modalities to bypass safeguards like prompt filters and post-hoc safety checkers, thus exposing and highlighting the vulnerabilities in existing defense mechanisms. Our codes are available at https://github.com/cure-lab/MMA-Diffusion.Keywords
All Related Versions
Funding Information
- National Key Research and Development Program of China (2021YFF0901503)
- National Natural Science Foundation of China (62206287)
This publication has 17 references indexed in Scilit:
- Unsafe Diffusion: On the Generation of Unsafe Images and Hateful Memes From Text-To-Image Models2023
- Ablating Concepts in Text-to-Image Diffusion Models2023
- Erasing Concepts from Diffusion Models2023
- Segment Anything2023
- RIATIG: Reliable and Imperceptible Adversarial Text-to-Image Generation with Natural Prompts2023
- Safe Latent Diffusion: Mitigating Inappropriate Degeneration in Diffusion Models2023
- A Pilot Study of Query-Free Adversarial Attack against Stable Diffusion2023
- Can Machines Help Us Answering Question 16 in Datasheets, and In Turn Reflecting on Inappropriate Content?2022
- High-Resolution Image Synthesis with Latent Diffusion Models2022
- Towards Evaluating the Robustness of Neural Networks2017