Summary of Contrasting Adversarial Perturbations: the Space Of Harmless Perturbations, by Lu Chen et al.
Contrasting Adversarial Perturbations: The Space of Harmless Perturbations
by Lu Chen, Shaofeng Li, Benhao Huang, Fan Yang, Zheng Li, Jie Li, Yuan Luo
First submitted to arxiv on: 3 Feb 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: None
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 research paper, the authors uncover a previously unknown aspect of deep neural networks’ behavior. Specifically, they show that certain harmless perturbations can be applied to inputs without changing the network’s output. This phenomenon arises from non-injective functions used within DNNs, which enable multiple distinct inputs to map to the same output. The authors then leverage this property to develop a family of general perturbation spaces that are redundant for the DNN’s decision-making process. These spaces can be used to hide sensitive data and serve as a means of model identification. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Imagine you’re trying to understand how deep learning models work, but you don’t have a technical background. In this paper, researchers found something cool! They discovered that certain tiny changes to the input data won’t affect what the model decides. This is because some types of math used in these models allow multiple inputs to look the same to the model. The authors then figured out how to use this property to hide important information and identify the type of model being used. |
Keywords
* Artificial intelligence * Deep learning
