Summary of Low-rank Adversarial Pgd Attack, by Dayana Savostianova et al.
Low-Rank Adversarial PGD Attack
by Dayana Savostianova, Emanuele Zangrando, Francesco Tudisco
First submitted to arxiv on: 16 Oct 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Numerical Analysis (math.NA); Machine Learning (stat.ML)
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 new method for generating effective adversarial attacks on deep neural networks is proposed, building upon the widely used Projected Gradient Descent (PGD) approach. By recognizing that PGD perturbations often primarily affect a subset of the image’s singular value spectrum, making them approximately low-rank, the authors develop a variation of PGD that efficiently computes low-rank attacks. This method is tested on various standard and robust models, demonstrating competitive performance to traditional full-rank PGD while requiring significantly less memory. The proposed approach has potential applications in adversarial training for improved model robustness. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary A team of researchers created a new way to make fake data that can trick artificial intelligence models. This method is based on an existing technique called Projected Gradient Descent (PGD), which is often used to test how well AI models work. By understanding that PGD can mostly affect certain parts of an image, the scientists developed a version of PGD that is faster and uses less memory while still being effective. They tested this new method on different types of AI models and found that it works just as well as the original PGD, but with some benefits. |
Keywords
» Artificial intelligence » Gradient descent
