Summary of Oslo: One-shot Label-only Membership Inference Attacks, by Yuefeng Peng et al.
OSLO: One-Shot Label-Only Membership Inference Attacks
by Yuefeng Peng, Jaechul Roh, Subhransu Maji, Amir Houmansadr
First submitted to arxiv on: 27 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 The proposed One-Shot Label-Only (OSLO) membership inference attack (MIA) can accurately determine whether a given sample is part of a target model’s training set with high precision, using just one query. This outperforms existing label-only attacks that require thousands of queries and achieve lower precisions. OSLO relies on transfer-based black-box adversarial attacks, leveraging the idea that member samples are more resistant to perturbations than non-members. The method is evaluated against state-of-the-art label-only MIA and demonstrates significant improvements in precision and true positive rate (TPR) under the same false positive rates (FPR), showcasing its effectiveness on various defense mechanisms. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary OSLO is a new way to figure out if a picture or data point came from a certain group of examples used to train a machine learning model. Normally, it takes many tries to get this information right, but OSLO can do it with just one try! This is important because it helps us understand how secure our models are and how well they’re protected from people trying to cheat or manipulate them. |
Keywords
» Artificial intelligence » Inference » Machine learning » One shot » Precision
