Summary of Hard-label Based Small Query Black-box Adversarial Attack, by Jeonghwan Park et al.
Hard-label based Small Query Black-box Adversarial Attack
by Jeonghwan Park, Paul Miller, Niall McLaughlin
First submitted to arxiv on: 9 Mar 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 The paper proposes a novel approach to hard-label-based black-box adversarial attacks, which leverages transferability between white-box surrogate models and target models. Unlike existing methods, this work focuses on hard labels and uses a pretrained surrogate model to guide the optimization process. The proposed method achieves significantly higher query efficiency and attack success rates compared to benchmarks, especially at small query budgets. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper develops a new way to make black-box AI systems vulnerable to attacks by using predictions from another AI system as hints. This helps to reduce the number of times an attacker needs to query the target AI model to succeed. The approach is tested on various types of AI models and shows that it can be much more effective than current methods, especially when only a limited number of queries are allowed. |
Keywords
* Artificial intelligence * Optimization * Transferability
