Privacy Amplification for Matrix Mechanisms
Privacy amplification exploits randomness in data selection to provide tighter differential privacy (DP) guarantees. This analysis is key to DP-SGD's success in machine learning, but, is not readily applicable to the newer state-of-the-art algorithms. This is because these algorithms, known as DP-FT...
Bewaard in:
| Hoofdauteurs: | , , , |
|---|---|
| Formaat: | Artikel |
| Gepubliceerd in: |
2023
|
| Onderwerpen: | |
| Online toegang: | https://arxiv.org/abs/2310.15526 http://arxiv.org/abs/2310.15526 |
| Tags: |
Geen labels, Wees de eerste die dit record labelt!
|