Saved in:
| Main Authors: | , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2412.08559 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866909630820515840 |
|---|---|
| author | Wei, Rongzhe Li, Mufei Ghassemi, Mohsen Kreačić, Eleonora Li, Yifan Yue, Xiang Li, Bo Potluru, Vamsi K. Li, Pan Chien, Eli |
| author_facet | Wei, Rongzhe Li, Mufei Ghassemi, Mohsen Kreačić, Eleonora Li, Yifan Yue, Xiang Li, Bo Potluru, Vamsi K. Li, Pan Chien, Eli |
| contents | Large Language Models (LLMs) embed sensitive, human-generated data, prompting the need for unlearning methods. Although certified unlearning offers strong privacy guarantees, its restrictive assumptions make it unsuitable for LLMs, giving rise to various heuristic approaches typically assessed through empirical evaluations. These standard evaluations randomly select data for removal, apply unlearning techniques, and use membership inference attacks (MIAs) to compare unlearned models against models retrained without the removed data. However, to ensure robust privacy protections for every data point, it is essential to account for scenarios in which certain data subsets face elevated risks. Prior research suggests that outliers, particularly including data tied to minority groups, often exhibit higher memorization propensity which indicates they may be more difficult to unlearn. Building on these insights, we introduce a complementary, minority-aware evaluation framework to highlight blind spots in existing frameworks. We substantiate our findings with carefully designed experiments, using canaries with personally identifiable information (PII) to represent these minority subsets and demonstrate that they suffer at least 20% higher privacy leakage across various unlearning methods, MIAs, datasets, and LLM scales. Our proposed minority-aware evaluation framework marks an essential step toward more equitable and comprehensive assessments of LLM unlearning efficacy. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2412_08559 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Underestimated Privacy Risks for Minority Populations in Large Language Model Unlearning Wei, Rongzhe Li, Mufei Ghassemi, Mohsen Kreačić, Eleonora Li, Yifan Yue, Xiang Li, Bo Potluru, Vamsi K. Li, Pan Chien, Eli Machine Learning Large Language Models (LLMs) embed sensitive, human-generated data, prompting the need for unlearning methods. Although certified unlearning offers strong privacy guarantees, its restrictive assumptions make it unsuitable for LLMs, giving rise to various heuristic approaches typically assessed through empirical evaluations. These standard evaluations randomly select data for removal, apply unlearning techniques, and use membership inference attacks (MIAs) to compare unlearned models against models retrained without the removed data. However, to ensure robust privacy protections for every data point, it is essential to account for scenarios in which certain data subsets face elevated risks. Prior research suggests that outliers, particularly including data tied to minority groups, often exhibit higher memorization propensity which indicates they may be more difficult to unlearn. Building on these insights, we introduce a complementary, minority-aware evaluation framework to highlight blind spots in existing frameworks. We substantiate our findings with carefully designed experiments, using canaries with personally identifiable information (PII) to represent these minority subsets and demonstrate that they suffer at least 20% higher privacy leakage across various unlearning methods, MIAs, datasets, and LLM scales. Our proposed minority-aware evaluation framework marks an essential step toward more equitable and comprehensive assessments of LLM unlearning efficacy. |
| title | Underestimated Privacy Risks for Minority Populations in Large Language Model Unlearning |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2412.08559 |