Summary of Beyond Full Poisoning: Effective Availability Attacks with Partial Perturbation, by Yu Zhe et al.
Beyond Full Poisoning: Effective Availability Attacks with Partial Perturbation
by Yu Zhe, Jun Sakuma
First submitted to arxiv on: 2 Jul 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Cryptography and Security (cs.CR); 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 proposed Parameter Matching Attack (PMA) is a novel availability attack approach that can cause more than a 30% performance drop even when only a portion of the training data can be perturbed. Unlike existing methods, PMA optimizes perturbations to make the trained model perform poorly, and experimental results across four datasets demonstrate its effectiveness in degrading model performance. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The researchers created a new way to protect data from being misused by making machine learning models perform badly when they’re trained on some messed-up data. This is called the Parameter Matching Attack (PMA). It’s special because it can make the models perform really poorly even if only part of the training data is changed. The experiment showed that PMA works better than other methods in degrading model performance. |
Keywords
* Artificial intelligence * Machine learning
