Skip to content

Fully Homomorphic Encryption for Machine Learning

Year: 2018

type: thesis collection: ppml status: auto

Summary

Auto-extracted summary (unreviewed)

This PhD thesis (Michele Minelli, ENS Paris, 2018) develops FHE techniques motivated by homomorphic inference on trained machine learning models. It proposes an FHE scheme tailored to evaluating neural networks on encrypted inputs whose cost is essentially independent of the number of layers, a low-overhead circuit-privacy technique for GSW-style FHE that hides the evaluated model, and an implemented homomorphic-encryption protocol (based on DGHV) for private information retrieval.

Extracted details

Field Value Source Confidence
contribution FHE framework for homomorphic evaluation of neural networks with complexity essentially independent of network depth llm 0.60
contribution Circuit-privacy technique for GSW homomorphic evaluation with small overhead and unchanged security parameters llm 0.60
contribution Protocol and implementation for private information retrieval based on an extended DGHV scheme llm 0.60
schemes GSW llm 0.60
schemes DGHV llm 0.60
schemes TFHE llm 0.60
schemes LWE llm 0.60
libraries TFHE llm 0.60
libraries Keras llm 0.60
models neural networks llm 0.60
tasks homomorphic neural network inference llm 0.60
tasks circuit privacy / model confidentiality llm 0.60
tasks private information retrieval llm 0.60
threat model user-server outsourcing; server does not hold the decryption key; circuit privacy protects the server's model from the user llm 0.60
interaction non-interactive FHE llm 0.60
code url https://github.com/tfhe/tfhe (URL found in the full text) regex 0.50
code url https://github.com/keras-team/keras (URL found in the full text) regex 0.50
code url https://github.com/shaih/HElib (URL found in the full text) regex 0.50
llm relevance medium llm 0.60

Keyword mentions

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

  • scheme: GSW 65 [p.10], Leveled / somewhat HE 31 [p.14], TFHE 21 [p.10], DGHV 20 [p.11], PIR 18 [p.5], MPC 16 [p.23], Oblivious transfer 13 [p.11], RSA 10 [p.17], Zero-knowledge proofs 9 [p.22], Paillier 8 [p.17], BGV 3 [p.50], Multi-key HE 3 [p.49], Threshold HE 3 [p.49], Garbled circuits 3 [p.63], Partially homomorphic encryption 3 [p.17], ElGamal 2 [p.17], Functional encryption 2 [p.137]
  • library: HElib 11 [p.50], Microsoft SEAL 4 [p.50], Concrete (Zama) 4 [p.11], TensorFlow 1 [p.147]
  • model: MLP 4 [p.10], CNN 3 [p.63], SVM 1 [p.148], Decision tree 1 [p.87]
  • dataset: MNIST 18 [p.10], Credit card / German credit 1 [p.106]
  • hardware: GPU 3 [p.59]
  • technique: Bootstrapping 115 [p.9], Key switching 15 [p.9], Sigmoid 9 [p.57], Non-interactive 7 [p.22], Convolution 6 [p.56], Softmax 5 [p.58], Modulus switching 3 [p.45], Semi-honest model 3 [p.21], Malicious model 3 [p.21], ReLU 2 [p.58], SIMD packing 2 [p.50], Two-party computation 2 [p.84]
  • assumption: Lattice-based 126 [p.9], LWE 119 [p.9], RLWE 17 [p.33], IND-CPA 12 [p.28]
  • task: Medical / healthcare 7 [p.16], Speech recognition 4 [p.59], Language modeling / generation 2 [p.30]
  • topic: Homomorphic encryption 222 [p.5], Benchmark 5 [p.11], Open source 4 [p.140]

Files

Path (under papers/) Collection Pages Size SHA-256 Note
references/ppml/theses-and-textbooks/minelli-2018_fully-homomorphic-encryption-for-machine-learning-phd-thesis.pdf ppml/theses-and-textbooks 152 3.2 MB 010b03cc3038 was Minelli et al. - Sciences Mathématiques de Paris Centre Spécialité Informatique.pdf

Database key: fully-homomorphic-encryption-for-machine-learning · created 2026-11-05 · updated 2026-11-05 · edit with python3 tools/paperdb.py set fully-homomorphic-encryption-for-machine-learning COLUMN VALUE