Datamodel Distance: A New Metric for Privacy
Proceedings of the AAAI Symposium Series · 08.11.2024 · 4 (1) · 68-75
DOI: 10.1609/aaaiss.v4i1.31773 ↗Kurz gefasst
Fachartikel in der Zeitschrift Proceedings of the AAAI Symposium Series, online erschienen am 08.11.2024. Verfasst von Paul Lintilhac und 2 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref).
Abstract (Originalsprache)
Recent work developing Membership Inference Attacks has demonstrated that certain points in the dataset are often in- trinsically easier to attack than others. In this paper, we intro- duce a new pointwise metric, the Datamodel Distance, and show that it is empirically correlated to and establishes a theoreti- cal lower bound for the success probability for a point under the LiRA Membership Inference Attack. This establishes a connection between the concepts of Datamodels and Member- ship Inference, and also gives new intuitive explanations for why certain points are more susceptible to attack than others. We then use datamodels as a lens through which to investigate the Privacy Onion Efect.
Zitieren
APA
Lintilhac, P., Scheible, H., & Bastian, N. D. (2024). Datamodel Distance: A New Metric for Privacy. Proceedings of the AAAI Symposium Series, 4(1), 68-75. https://doi.org/10.1609/aaaiss.v4i1.31773
Vancouver
Lintilhac P, Scheible H, Bastian ND. Datamodel Distance: A New Metric for Privacy. Proceedings of the AAAI Symposium Series. 2024;4(1):68-75. doi:10.1609/aaaiss.v4i1.31773
BibTeX
@article{Lintilhac2024,
author = {Lintilhac, Paul and Scheible, Henry and Bastian, Nathaniel D.},
title = {Datamodel Distance: A New Metric for Privacy},
journal = {Proceedings of the AAAI Symposium Series},
year = {2024},
volume = {4},
number = {1},
pages = {68--75},
doi = {10.1609/aaaiss.v4i1.31773}
}
EUVIMED