DOI | Trouver le DOI : https://doi.org/10.1007/978-981-99-9785-5_28 |
---|
Auteur | Rechercher : Xue, LiangIdentifiant ORCID : https://orcid.org/0000-0001-8069-3182; Rechercher : Lin, XiaodongIdentifiant ORCID : https://orcid.org/0000-0001-8916-6645; Rechercher : Xiong, Pulei1Identifiant ORCID : https://orcid.org/0000-0002-3460-6946 |
---|
Affiliation | - Conseil national de recherches du Canada. Technologies numériques
|
---|
Format | Texte, Article |
---|
Conférence | First International Conference on Artificial Intelligence Security and Privacy, AIS&P, December 3–5, 2023, Guangzhou, China |
---|
Sujet | machine learning; Ddcision tree evaluation; privacy preservation |
---|
Résumé | Machine learning enables organizations and individuals to improve efficiency and productivity. With an abundance of data and computational resources, large companies can build complex machine learning models and provide prediction services to clients. One example is decision tree evaluation, where a client can access the trained decision tree model with its input and obtain the classification result. However, the privacy issues on model parameters and clients’ inputs and results need to be addressed. In this paper, we propose a privacy-preserving decision tree evaluation scheme, where we first design an improved interval encoding method that can hide parameters representing an interval. Then, based on the interval encoding method, hash functions, and the Diffie-Hellman key agreement technique, a model owner can generate a set of encodings for the decision tree model and send them to a client, who can determine the classification result based on its input and the encodings. The proposed scheme conceals the model parameters from clients and preserves the data privacy of clients, and only one round of communication between the two entities is needed. We provide a formal security proof that demonstrates the privacy preservation property of our scheme. Performance evaluation shows the practicability of the proposed scheme. |
---|
Date de publication | 2024 |
---|
Maison d’édition | Springer Nature |
---|
Dans | |
---|
Série | |
---|
Langue | anglais |
---|
Publications évaluées par des pairs | Oui |
---|
Exporter la notice | Exporter en format RIS |
---|
Signaler une correction | Signaler une correction (s'ouvre dans un nouvel onglet) |
---|
Identificateur de l’enregistrement | 0f568893-ce9b-4105-adc0-e89f02fafef7 |
---|
Enregistrement créé | 2024-02-27 |
---|
Enregistrement modifié | 2024-02-28 |
---|