Privacy-Preserving Machine Learning Using Cryptography

Christian Rechberger, Roman Walch*

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingChapterpeer-review


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.

Original languageEnglish
Title of host publicationSecurity and Artificial Intelligence
Subtitle of host publicationA Crossdisciplinary Approach
Place of PublicationCham
PublisherSpringer Science and Business Media Deutschland GmbH
Number of pages21
ISBN (Electronic)978-3-030-98795-4
ISBN (Print)978-3-030-98794-7
Publication statusPublished - 2022

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume13049 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349


  • FHE
  • Machine learning
  • MPC
  • Privacy

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)


Dive into the research topics of 'Privacy-Preserving Machine Learning Using Cryptography'. Together they form a unique fingerprint.

Cite this