TY - CHAP
T1 - Privacy-Preserving Machine Learning Using Cryptography
AU - Rechberger, Christian
AU - Walch, Roman
N1 - Funding Information:
Acknowledgments. This work was supported by the “DDAI” COMET Module within the COMET – Competence Centers for Excellent Technologies Programme, funded by the Austrian Federal Ministry for Transport, Innovation and Technology (bmvit), the Austrian Federal Ministry for Digital and Economic Affairs (bmdw), the Austrian Research Promotion Agency (FFG), the province of Styria (SFG) and partners from industry and academia. The COMET Programme is managed by FFG.
Publisher Copyright:
© 2022, Springer Nature Switzerland AG.
PY - 2022
Y1 - 2022
N2 - Data scientists require an extensive training set to train an accurate and reliable machine learning model – the bigger and diverse the training set, the better. However, acquiring such a vast training set can be difficult, especially when sensitive user data is involved. The General Data Protection Regulation (GDPR) and similar regulations may prohibit the gathering and processing of this sensitive data. Privacy-preserving cryptographic protocols and primitives, like secure multi-party computation (MPC) and fully homomorphic encryption (FHE), may provide a solution to this problem. They allow us to perform calculations on private and unknown data and can, therefore, be used to classify and train on GDPR protected data sets. While still considered very inefficient, privacy-preserving machine learning using MPC and FHE has been heavily researched in recent years. In this chapter, we give an introduction to MPC and FHE, how they can be used, their limitations, and describe how state-of-the-art publications apply them to machine learning algorithms.
AB - Data scientists require an extensive training set to train an accurate and reliable machine learning model – the bigger and diverse the training set, the better. However, acquiring such a vast training set can be difficult, especially when sensitive user data is involved. The General Data Protection Regulation (GDPR) and similar regulations may prohibit the gathering and processing of this sensitive data. Privacy-preserving cryptographic protocols and primitives, like secure multi-party computation (MPC) and fully homomorphic encryption (FHE), may provide a solution to this problem. They allow us to perform calculations on private and unknown data and can, therefore, be used to classify and train on GDPR protected data sets. While still considered very inefficient, privacy-preserving machine learning using MPC and FHE has been heavily researched in recent years. In this chapter, we give an introduction to MPC and FHE, how they can be used, their limitations, and describe how state-of-the-art publications apply them to machine learning algorithms.
KW - FHE
KW - Machine learning
KW - MPC
KW - Privacy
UR - http://www.scopus.com/inward/record.url?scp=85128023017&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-98795-4_6
DO - 10.1007/978-3-030-98795-4_6
M3 - Chapter
AN - SCOPUS:85128023017
SN - 978-3-030-98794-7
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 109
EP - 129
BT - Security and Artificial Intelligence
PB - Springer Science and Business Media Deutschland GmbH
CY - Cham
ER -