Submit

Login Login

Register Register

Home / Journals / CMC / Online First / doi:10.32604/cmc.2026.087848
Journal Menu
Special Issues
Table of Content

All issues

Online First

2026

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

Open Access

REVIEW

Privacy-Preserving Cryptography for Machine Learning: A Comprehensive Review of Homomorphic Encryption, Differential Privacy, Secure Computation, and Post-Quantum Foundations

Quang-Vinh Dang1, Dat Le2,*, Minh Ngoc Dinh3, Ngoc-Son-An Nguyen4
1 School of Computing and Innovative Technologies, British University Vietnam, Hung Yen, Hanoi, Vietnam
2 Department of Science, Technology and International Projects, University of Economics and Finance, Ho Chi Minh City, Vietnam
3 School of Data Science and Information Technology, Millennia Education, Ho Chi Minh City, Vietnam
4 Faculty of Information Technology, Industrial University of Ho Chi Minh City, Ho Chi Minh City, Vietnam
* Corresponding Author: Dat Le. Email: email
(This article belongs to the Special Issue: Privacy-Preserving AI: Encryption and Differential Privacy)

Computers, Materials & Continua https://doi.org/10.32604/cmc.2026.087848

Received 24 June 2026; Accepted 25 August 2026; Published online 10 September 2026

Abstract

Deploying machine learning on sensitive data has made privacy a first-order design constraint. Over the past five years privacy-preserving machine learning (PPML) has matured from isolated proofs of concept into an ecosystem of cryptographic and statistical techniques, each occupying a distinct point in the trade-off space among confidentiality, integrity, utility, cost and trust. This review surveys that ecosystem for a research audience. We examine homomorphic encryption, including several currently available bootstrapping styles and encrypted transformer inference; differential privacy in deep learning, analytics and the private fine-tuning of large language models; privacy budget management through composition theorems, numerical accountants and empirical auditing; secure multi-party computation; zero-knowledge proofs as an integrity complement to confidentiality; federated learning, covering secure aggregation, gradient-leakage attacks, verifiable aggregation and federated adaptation of large models; encrypted search and its leakage-abuse attacks; and post-quantum foundations following the 2024 standardization of Module-Lattice-Based Key-Encapsulation Mechanism (ML-KEM), Module-Lattice-Based Digital Signature Algorithm (ML-DSA) and Stateless Hash-Based Digital Signature Algorithm (SLH-DSA). Beyond surveying these areas, we contribute three syntheses intended to make the material actionable: an analysis of compositional compatibility identifying where combining techniques creates conflicts rather than complementarity; a scenario-driven selection guide mapping deployment situations to technique combinations; and an examination of the gap between research prototypes and production systems spanning engineering complexity, key management, regulatory compliance and evidence from healthcare, finance and smart-city deployments. We close by identifying the open problems that will shape the next phase of research.

Keywords

Privacy-preserving machine learning; homomorphic encryption; differential privacy; secure multi-party computation; federated learning; zero-knowledge proofs; searchable encryption; post-quantum cryptography; privacy budget; privacy auditing; large language models

  • 136

    View

  • 21

    Download

  • 0

    Like

Related articles
Copyright© 2026 Tech Science Press
© 1997-2026 TSP (Henderson, USA) unless otherwise stated

Share Link