Summary of Invisible Backdoor Attacks on Diffusion Models, by Sen Li et al.
Invisible Backdoor Attacks on Diffusion Models
by Sen Li, Junchi Ma, Minhao Cheng
First submitted to arxiv on: 2 Jun 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Cryptography and Security (cs.CR); Computer Vision and Pattern Recognition (cs.CV)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
Summary difficulty | Written by | Summary |
---|---|---|
High | Paper authors | High Difficulty Summary Read the original abstract here |
Medium | GrooveSquid.com (original content) | Medium Difficulty Summary A novel optimization framework for diffusion models is presented, which enables the creation of invisible triggers that enhance the stealthiness and resilience of backdoors in both unconditional and conditional diffusion models. The proposed framework demonstrates the backdooring of text-guided image editing and inpainting pipelines, as well as its application to model watermarking for ownership verification. Experimental results on various samplers and datasets show the efficacy and stealthiness of the approach. |
Low | GrooveSquid.com (original content) | Low Difficulty Summary Diffusion models can generate high-quality images, but they are vulnerable to malicious attacks. A new way to make these attacks more difficult to detect is by creating invisible triggers that can manipulate the generated images. This technique is useful for both unconditional and conditional diffusion models, and it also helps protect model ownership by watermarking the generated images. |
Keywords
» Artificial intelligence » Diffusion » Optimization