Summary of Sok: Analyzing Adversarial Examples: a Framework to Study Adversary Knowledge, by Lucas Fenaux and Florian Kerschbaum
SoK: Analyzing Adversarial Examples: A Framework to Study Adversary Knowledge
by Lucas Fenaux, Florian Kerschbaum
First submitted to arxiv on: 22 Feb 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: 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 research paper, the authors investigate the concept of adversarial examples in machine learning, specifically in the context of image classification. They identify a lack of formalization and study of adversary knowledge when mounting attacks, which has led to a complex and hard-to-compare space of attack research. The authors propose a theoretical framework inspired by order theory to standardize attacks and present an adversarial example game, similar to cryptographic games, to facilitate comparison. They survey recent attacks in the image classification domain and classify their adversary’s knowledge using this framework, revealing new insights on the difficulty associated with white-box and transferable threat models. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Adversarial examples are like sneaky tricks that can fool machine learning models into making mistakes. In this study, scientists found that people haven’t been very good at understanding how these trick attacks work. They created a special way to look at this problem using math ideas from order theory. Then, they designed a game to test different attack methods and looked at what others have done in the past. This helped them figure out some new things about how hard it is to make models do silly things. |
Keywords
* Artificial intelligence * Image classification * Machine learning
