Summary of An Invisible Backdoor Attack Based on Semantic Feature, by Yangming Chen
An Invisible Backdoor Attack Based On Semantic Feature
by Yangming Chen
First submitted to arxiv on: 19 May 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI)
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 proposes a novel backdoor attack that makes imperceptible changes to deep neural network (DNN) models, allowing them to behave normally on benign samples while making wrong predictions for samples containing triggers. The attack uses pre-trained victim models to extract semantic features from clean images and generates trigger patterns associated with high-level features using channel attention. The results demonstrate the effectiveness of the attack in achieving high success rates across three prominent image classification DNNs on three standard datasets, while maintaining robustness against backdoor defenses. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper introduces a new way to trick deep learning models, making them make mistakes without anyone noticing. The authors use a clever method to create “triggers” that can be hidden in images, allowing the model to behave normally until it sees one of these triggers. They tested this attack on popular image classification models and found that it was very effective. This could have important implications for how we use deep learning models in real-world applications. |
Keywords
» Artificial intelligence » Attention » Deep learning » Image classification » Neural network
