Summary of Backdoor Attacks Against Image-to-image Networks, by Wenbo Jiang and Hongwei Li and Jiaming He and Rui Zhang and Guowen Xu and Tianwei Zhang and Rongxing Lu
Backdoor Attacks against Image-to-Image Networks
by Wenbo Jiang, Hongwei Li, Jiaming He, Rui Zhang, Guowen Xu, Tianwei Zhang, Rongxing Lu
First submitted to arxiv on: 15 Jul 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
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 In this paper, researchers investigate the susceptibility of Image-to-Image (I2I) networks to backdoor attacks. Despite their impressive performance in tasks like image super-resolution and denoising, I2I networks are vulnerable to malicious input images containing a trigger that causes the network to output a predefined image. The authors propose a novel attack technique using targeted universal adversarial perturbations (UAPs) as triggers. They also develop a multi-task learning approach with dynamic weighting methods for accelerating convergence rates in the backdoor training process. Experimental results demonstrate the effectiveness of this I2I backdoor on state-of-the-art network architectures and its robustness against different defense mechanisms. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper explores how Image-to-Image (I2I) networks can be tricked into producing wrong output images. The authors created a special type of attack that makes these networks behave normally with normal input, but produce a specific image when given a trigger. They developed a way to create this trigger using a mathematical formula called universal adversarial perturbations. The team tested their attack on several different I2I networks and found it worked well. This has implications for how we use these types of networks in real-world applications. |
Keywords
» Artificial intelligence » Multi task » Super resolution
