Summary of Backdoor Attack Against Vision Transformers Via Attention Gradient-based Image Erosion, by Ji Guo et al.
Backdoor Attack Against Vision Transformers via Attention Gradient-Based Image Erosion
by Ji Guo, Hongwei Li, Wenbo Jiang, Guoming Lu
First submitted to arxiv on: 30 Oct 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 The paper presents a study on Vision Transformers (ViTs) in computer vision tasks, which outperform traditional Convolutional Neural Networks (CNNs). However, like CNNs, ViTs are susceptible to backdoor attacks that manipulate the model’s predictions. The existing backdoor attacks against ViTs have a limitation in balancing stealthiness and effectiveness. To address this issue, researchers propose an innovative approach that optimizes the attack strategy to achieve both stealthy and effective backdoor attacks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Backdoor attacks are a kind of cyber threat where hackers secretly alter a model’s predictions for certain inputs. This paper talks about how these attacks work on special computer models called Vision Transformers (ViTs). Right now, some clever ways have been found to trick ViTs into making wrong choices, but the problem is that these tricks don’t quite hit the right balance between being sneaky and doing damage. |
