Summary of Hidden in Plain Sight: Undetectable Adversarial Bias Attacks on Vulnerable Patient Populations, by Pranav Kulkarni et al.
Hidden in Plain Sight: Undetectable Adversarial Bias Attacks on Vulnerable Patient Populations
by Pranav Kulkarni, Andrew Chan, Nithya Navarathna, Skylar Chan, Paul H. Yi, Vishwa S. Parekh
First submitted to arxiv on: 8 Feb 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); 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 This AI research paper proposes a novel approach to studying and mitigating biases in deep learning (DL) models used in medical imaging. The authors investigate demographically targeted label poisoning attacks, which can introduce undetectable underdiagnosis bias in DL models. Their results show that these attacks are highly selective for bias in the targeted group, degrading model performance without impacting overall performance. Furthermore, the paper finds that biased DL models can propagate prediction bias even when evaluated with external datasets. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary AI is getting better at reading medical images like X-rays and MRIs, but it’s not always fair. This paper looks into how to make sure AI doesn’t unfairly treat certain groups of people. They found a way to trick the AI into making mistakes by giving it fake information about some patients. The results show that this “trick” works better for certain groups than others. This means that the AI might not be fair even when it’s tested with new images. |
Keywords
* Artificial intelligence * Deep learning
