Skip to content

Recent advances of privacy-preserving machine learning based on (Fully) Homomorphic Encryption

Authors: Cheng Hong
Year: 2025
Venue: Security and Safety

type: survey collection: ppml status: auto

Links: DOI

Summary

Auto-extracted summary (unreviewed)

This short survey article organises FHE schemes into Gentry's four generations (plus mixed-mode FHE) and reviews representative FHE-based privacy-preserving machine learning results so that practitioners can pick a suitable scheme and gauge expected efficiency for training or inference of a given model. It concludes that FHE is already practical for small and moderate PPML tasks, that each scheme generation has complementary strengths, and that hardware-software co-optimisation is the key to reaching complex tasks such as ImageNet classification.

Abstract

Fully Homomorphic Encryption (FHE), known for its ability to process encrypted data without decryption, is a promising technique for solving privacy concerns in the machine learning era. However, there are many kinds of available FHE schemes and way more FHE-based solutions in the literature, and they are still fast evolving, making it difficult to get a complete view. This article aims to introduce recent representative results of FHE-based privacy-preserving machine learning, helping users understand the pros and cons of different kinds of solutions, and choose an appropriate approach for their needs.
Source: crossref

Extracted details

Field Value Source Confidence
contribution overview of FHE scheme generations and representative FHE-based PPML applications llm 0.70
contribution guidance on pros and cons of different FHE-based PPML solutions for choosing an approach llm 0.70
contribution assessment of maturity and expected efficiency of FHE-based PPML llm 0.70
schemes BFV llm 0.70
schemes BGV llm 0.70
schemes FHEW llm 0.70
schemes TFHE llm 0.70
schemes CKKS llm 0.70
schemes DGHV llm 0.70
tasks privacy-preserving machine learning llm 0.70
tasks private training llm 0.70
tasks private inference llm 0.70
interaction non-interactive FHE llm 0.70
limitations FHE is practical only for small or moderate PPML tasks; complex tasks such as ImageNet classification remain far off llm 0.70
limitations no single FHE generation can replace the others llm 0.70
llm relevance medium llm 0.70

Keyword mentions

Counts of vocabulary terms in the full text (first page of occurrence in brackets).

  • scheme: CKKS 18 [p.2], TFHE 12 [p.2], BFV 5 [p.2], BGV 4 [p.2], Leveled / somewhat HE 3 [p.3], FHEW 2 [p.5], MPC 2 [p.2]
  • library: Zama 2 [p.3], EzPC / CrypTFlow 1 [p.6]
  • model: Logistic regression 7 [p.4], CNN 6 [p.2], ResNet 4 [p.4], Decision tree 3 [p.4], Transformer 2 [p.4], Random forest 1 [p.4], Gradient boosting / XGBoost 1 [p.6], k-means 1 [p.5]
  • dataset: MNIST 4 [p.3]
  • hardware: FPGA 4 [p.5], GPU 4 [p.5], NVIDIA A100 1 [p.5], ASIC 1 [p.5]
  • technique: Bootstrapping 15 [p.2], Programmable bootstrapping 10 [p.3], ReLU 8 [p.2], SIMD packing 6 [p.2], Convolution 5 [p.2], Sigmoid 4 [p.3], Quantization 3 [p.3], Lookup table 2 [p.3], Non-interactive 2 [p.2], Two-party computation 2 [p.6]
  • assumption: RLWE 2 [p.2], LWE 2 [p.5]
  • task: Genomics 6 [p.4]
  • topic: Homomorphic encryption 71 [p.1], Privacy-preserving machine learning 18 [p.1], Private / secure inference 2 [p.3]

Files

Path (under papers/) Collection Pages Size SHA-256 Note
references/ppml/surveys/hong-2024_recent-advances-of-privacy-preserving-ml-based-on-fhe.pdf ppml/surveys 7 908.7 KB 7b1d6a51057a was Hong - 2024 - Recent advances of privacy-preserving machine learning based on (Fully) Homomorphic Encryption.pdf

Database key: recent-advances-of-privacy-preserving-machine-learning-based-on-fully · created 2026-11-05 · updated 2026-11-05 · edit with python3 tools/paperdb.py set recent-advances-of-privacy-preserving-machine-learning-based-on-fully COLUMN VALUE