Duddu, Vasisht ; Das, Anudeep ; Khayata, Nora ; Yalame, Hossein ; Schneider, Thomas ; Asokan, N. (2024)
Attesting Distributional Properties of Training Data for Machine Learning.
29th European Symposium on Research in Computer Security. Bydgoszcz, Poland (16.09.2024 -20.09.2024)
doi: 10.1007/978-3-031-70879-4_1
Konferenzveröffentlichung, Bibliographie
Kurzbeschreibung (Abstract)
The success of machine learning (ML) has been accompanied by increased concerns about its trustworthiness. Several jurisdictions are preparing ML regulatory frameworks. One such concern is ensuring that model training data has desirable distributional properties for certain sensitive attributes. For example, draft regulations indicate that model trainers are required to show that training datasets have specific distributional properties, such as reflecting the diversity of the population. We propose the novel notion of ML property attestation allowing a prover (e.g., model trainer) to demonstrate relevant properties of an ML model to a verifier (e.g., a customer) while preserving the confidentiality of sensitive data. We focus on the attestation of distributional properties of training data without revealing the data. We present an effective hybrid property attestation combining property inference with cryptographic mechanisms.
| Typ des Eintrags: | Konferenzveröffentlichung |
|---|---|
| Erschienen: | 2024 |
| Autor(en): | Duddu, Vasisht ; Das, Anudeep ; Khayata, Nora ; Yalame, Hossein ; Schneider, Thomas ; Asokan, N. |
| Art des Eintrags: | Bibliographie |
| Titel: | Attesting Distributional Properties of Training Data for Machine Learning |
| Sprache: | Englisch |
| Publikationsjahr: | 5 September 2024 |
| Verlag: | Springer |
| Buchtitel: | Computer Security - ESORICS 2024 |
| Reihe: | Lecture Notes in Computer Science |
| Band einer Reihe: | 14982 |
| Veranstaltungstitel: | 29th European Symposium on Research in Computer Security |
| Veranstaltungsort: | Bydgoszcz, Poland |
| Veranstaltungsdatum: | 16.09.2024 -20.09.2024 |
| DOI: | 10.1007/978-3-031-70879-4_1 |
| Kurzbeschreibung (Abstract): | The success of machine learning (ML) has been accompanied by increased concerns about its trustworthiness. Several jurisdictions are preparing ML regulatory frameworks. One such concern is ensuring that model training data has desirable distributional properties for certain sensitive attributes. For example, draft regulations indicate that model trainers are required to show that training datasets have specific distributional properties, such as reflecting the diversity of the population. We propose the novel notion of ML property attestation allowing a prover (e.g., model trainer) to demonstrate relevant properties of an ML model to a verifier (e.g., a customer) while preserving the confidentiality of sensitive data. We focus on the attestation of distributional properties of training data without revealing the data. We present an effective hybrid property attestation combining property inference with cryptographic mechanisms. |
| Fachbereich(e)/-gebiet(e): | 20 Fachbereich Informatik 20 Fachbereich Informatik > Praktische Kryptographie und Privatheit DFG-Sonderforschungsbereiche (inkl. Transregio) DFG-Sonderforschungsbereiche (inkl. Transregio) > Sonderforschungsbereiche DFG-Graduiertenkollegs DFG-Sonderforschungsbereiche (inkl. Transregio) > Sonderforschungsbereiche > SFB 1119: CROSSING – Kryptographiebasierte Sicherheitslösungen als Grundlage für Vertrauen in heutigen und zukünftigen IT-Systemen |
| Hinterlegungsdatum: | 29 Okt 2024 13:25 |
| Letzte Änderung: | 29 Okt 2024 13:27 |
| PPN: | |
| Export: | |
| Suche nach Titel in: | TUfind oder in Google |
 |
Frage zum Eintrag |
Optionen (nur für Redakteure)
 |
Redaktionelle Details anzeigen |
Drucken
Impressum