Practical Considerations of Fully Homomorphic Encryption in Privacy-Preserving Machine Learning¶
Authors: Dan Chia-Tien Lo; Yong Shi; Hossain Shahriar; Bobin Deng; Xinyue Zhang; Mei-Lan Chen
Year: 2024
Venue: 2024 IEEE International Conference on Big Data (BigData)
type: other collection: ppml status: auto
Links: DOI
Summary¶
Auto-extracted summary (unreviewed)
A short experience paper reporting experiments on training machine learning models over FHE-encrypted data in a client/server setup where the server never decrypts. The authors conclude that FHE preserves privacy but significantly degrades training performance and requires complex parameter tuning; SEAL and Concrete ML are the frameworks discussed.
Abstract¶
Machine learning has been successfully applied to big data analytics across various disciplines. However, as data is collected from diverse sectors, much of it is private and confidential. At the same time, one of the major challenges in machine learning is the slow training speed of large models, which often requires high-performance servers or cloud services. To protect data privacy while still allowing model training on such servers, privacy-preserving machine learning using Fully Homomorphic Encryption (FHE) has gained significant attention. However, its widespread adoption is hindered by performance degradation. This paper presents our experiments on training models over encrypted data using FHE. The results show that while FHE ensures privacy, it can significantly degrade performance, requiring complex tuning to optimize.
Source: pdf
Extracted details¶
| Field | Value | Source | Confidence |
|---|---|---|---|
| contribution | Experiments on FHE-based training over encrypted data | llm | 0.40 |
| contribution | Discussion of performance degradation and tuning burden of FHE for PPML | llm | 0.40 |
| libraries | Concrete ML | llm | 0.40 |
| libraries | SEAL | llm | 0.40 |
| tasks | privacy-preserving model training on encrypted data | llm | 0.40 |
| threat model | client/server; server computes on ciphertexts without the private key | llm | 0.40 |
| interaction | non-interactive FHE | llm | 0.40 |
| code url | https://github.com/Microsoft/SEAL (URL found in the full text) | regex | 0.50 |
| code url | https://github.com/zama-ai/concrete-ml (URL found in the full text) | regex | 0.50 |
| limitations | FHE significantly degrades training performance | llm | 0.40 |
| limitations | Requires complex tuning to optimize | llm | 0.40 |
| llm relevance | low | llm | 0.40 |
Keyword mentions¶
Counts of vocabulary terms in the full text (first page of occurrence in brackets).
- scheme: CKKS 7 [p.3], Differential privacy 7 [p.3], BFV 3 [p.3], BGV 2 [p.3]
- library: Concrete (Zama) 9 [p.3], Concrete ML 8 [p.3], Microsoft SEAL 6 [p.3], TenSEAL 3 [p.3], Zama 1 [p.6], PyTorch 1 [p.3]
- model: Logistic regression 10 [p.2], Random forest 6 [p.4], Gradient boosting / XGBoost 3 [p.4], SVM 2 [p.5], k-NN 2 [p.5], Decision tree 1 [p.4]
- dataset: Breast cancer (Wisconsin) 6 [p.3]
- hardware: TPU 1 [p.5]
- technique: Quantization 5 [p.5], Sigmoid 3 [p.3], Matrix multiplication 2 [p.1]
- task: Medical / healthcare 3 [p.6]
- topic: Homomorphic encryption 44 [p.1], Privacy-preserving machine learning 9 [p.1], Open source 3 [p.3]
Files¶
| Path (under papers/) | Collection | Pages | Size | SHA-256 | Note |
|---|---|---|---|---|---|
references/ppml/surveys/lo-2024_practical-considerations-of-fhe-in-privacy-preserving-ml.pdf |
ppml/surveys | 6 | 383.9 KB | 599f745dc254 |
was Lo et al. - 2024 - Practical Considerations of Fully Homomorphic Encryption in Privacy-Preserving Machine Learning.pdf |
Database key: practical-considerations-of-fully-homomorphic-encryption-in-privacy · created 2026-11-05 · updated 2026-11-05 · edit with python3 tools/paperdb.py set practical-considerations-of-fully-homomorphic-encryption-in-privacy COLUMN VALUE