Summary of Cross-input Certified Training For Universal Perturbations, by Changming Xu et al.
Cross-Input Certified Training for Universal Perturbations
by Changming Xu, Gagandeep Singh
First submitted to arxiv on: 15 May 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 A novel approach for trustworthy machine learning is proposed in this paper, which addresses the limitations of existing methods that focus primarily on single-input adversarial perturbations. The authors introduce CITRUS, a method for certified training of networks robust against universal adversarial perturbations (UAPs), and demonstrate its effectiveness through extensive evaluation across various datasets, architectures, and perturbation magnitudes. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper introduces a new way to make machine learning models more reliable and secure. The problem it solves is that many real-world attacks use input-agnostic methods, not just single-input ones. Current methods for training models are good at dealing with small changes in the data but don’t work well when there’s a big change. The authors show that their new method, called CITRUS, does better than current methods on both normal accuracy and how well it works against UAP attacks. |
Keywords
» Artificial intelligence » Machine learning
