Homomorphic Encryption for Machine Learning and Artificial Intelligence Applications¶
Year: 2022
type: application collection: ppml status: auto
Summary¶
Auto-extracted summary (unreviewed)
An Argonne National Laboratory technical report (ANL/NSE-22/54) that applies Principal Component Analysis and small fully connected neural networks to encrypted data using Microsoft SEAL's CKKS implementation, motivated by outsourcing reactor structural health monitoring and nondestructive evaluation analysis to third parties. MNIST serves as a proof of concept, with ReLU replaced by a polynomial approximation and results covering test accuracy, ciphertext storage size and timing, though no concrete numbers appear in the excerpt.
Extracted details¶
| Field | Value | Source | Confidence |
|---|---|---|---|
| contribution | PCA and fully connected neural network inference on CKKS-encrypted data using Microsoft SEAL | llm | 0.55 |
| contribution | proof-of-concept on MNIST with accuracy, storage-size and timing analysis | llm | 0.55 |
| contribution | polynomial approximation of ReLU for encrypted evaluation | llm | 0.55 |
| schemes | CKKS | llm | 0.55 |
| libraries | SEAL | llm | 0.55 |
| models | PCA | llm | 0.55 |
| models | fully connected neural network | llm | 0.55 |
| models | simple CNN | llm | 0.55 |
| datasets | MNIST | llm | 0.55 |
| tasks | image classification | llm | 0.55 |
| tasks | reactor structural health monitoring and nondestructive evaluation (motivating application) | llm | 0.55 |
| threat model | untrusted third-party analyst / server (client-server) | llm | 0.55 |
| interaction | non-interactive FHE | llm | 0.55 |
| llm relevance | medium | llm | 0.55 |
Keyword mentions¶
Counts of vocabulary terms in the full text (first page of occurrence in brackets).
- scheme: CKKS 4 [p.6]
- library: Microsoft SEAL 3 [p.6]
- model: CNN 12 [p.5]
- dataset: MNIST 8 [p.5]
- technique: ReLU 10 [p.5], Convolution 10 [p.5], Polynomial approximation 8 [p.5], Softmax 5 [p.9], Rescaling 2 [p.8]
- topic: Homomorphic encryption 25 [p.6]
Files¶
| Path (under papers/) | Collection | Pages | Size | SHA-256 | Note |
|---|---|---|---|---|---|
references/ppml/surveys/arnold-2022_homomorphic-encryption-for-machine-learning-and-ai-applications.pdf |
ppml/surveys | 17 | 915.8 KB | 0d0e86f54e88 |
was Arnold et al. - 2022 - Homomorphic Encryption for Machine Learning and Artificial Intelligence Applications.pdf |
Database key: homomorphic-encryption-for-machine-learning-and-artificial-intelligence · created 2026-11-05 · updated 2026-11-05 · edit with python3 tools/paperdb.py set homomorphic-encryption-for-machine-learning-and-artificial-intelligence COLUMN VALUE