Skip to content

Evaluation of Privacy-aware Support Vector Machine (SVM) Learning using Homomorphic Encryption

Authors: William J Buchanan; Hisham Ali
Year: 2025
Venue: arXiv preprint

type: benchmark collection: ppml status: auto

Links: arXiv:2503.04652

Summary

Auto-extracted summary (unreviewed)

This arXiv preprint measures the overhead of running scikit-learn-trained SVM models (linear and polynomial kernels) on homomorphically encrypted inputs with the OpenFHE library from Python. It sweeps multiplication depth, scale size, first modulus size, security level (128/192/256-bit), batch size and ring dimension, and concludes that ring dimension and modulus size are the two parameters that dominate performance, while SVM-Poly and SVM-Linear behave similarly.

Abstract

The requirement for privacy-aware machine learning increases as we continue to use PII (Personally Identifiable Information) within machine training. To overcome these privacy issues, we can apply Fully Homomorphic Encryption (FHE) to encrypt data before it is fed into a machine learning model. This involves creating a homomorphic encryption key pair, and where the associated public key will be used to encrypt the input data, and the private key will decrypt the output. But, there is often a performance hit when we use homomorphic encryption, and so this paper evaluates the performance overhead of using the SVM machine learning technique with the OpenFHE homomorphic encryption library. This uses Python and the scikit-learn library for its implementation. The experiments include a range of variables such as multiplication depth, scale size, first modulus size, security level, batch size, and ring dimension, along with two different SVM models, SVM-Poly and SVM-Linear. Overall, the results show that the two main parameters which affect performance are the ring dimension and the modulus size, and that SVM-Poly and SVM-Linear show similar performance levels.
Source: arxiv

Extracted details

Field Value Source Confidence
contribution FHE-based SVM inference (SVM-Linear and SVM-Poly) using an SVM model extracted from scikit-learn llm 0.70
contribution parameter sweep over multiplication depth, scale size, first modulus size, security level, batch size and ring dimension in OpenFHE llm 0.70
contribution finding that ring dimension and modulus size dominate performance overhead llm 0.70
schemes CKKS llm 0.70
libraries OpenFHE llm 0.70
libraries scikit-learn llm 0.70
models SVM (linear kernel) llm 0.70
models SVM (polynomial kernel, degree 3) llm 0.70
datasets credit approval llm 0.70
tasks privacy-preserving SVM inference llm 0.70
tasks binary classification llm 0.70
security level 128-bit, 192-bit and 256-bit configurations evaluated llm 0.70
threat model outsourced computation; data processor holds only the public key and never the secret key llm 0.70
interaction non-interactive FHE llm 0.70
code url https://github.com/openfheorg/openfhe-genomic-examples/blob/main/demo-chi2.cpp (URL found in the full text) regex 0.50
code url https://github.com/openfheorg/openfhe-development.2 (URL found in the full text) regex 0.50
code url https://github.com/openfheorg/openfhe-development (URL found in the full text) regex 0.50
limitations homomorphic encryption introduces a processing overhead llm 0.70
llm relevance low llm 0.70

Keyword mentions

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

  • scheme: CKKS 32 [p.2], BFV 10 [p.2], BGV 9 [p.2], TFHE 8 [p.2], MPC 3 [p.16], Partially homomorphic encryption 3 [p.1], FHEW 2 [p.2]
  • library: OpenFHE 17 [p.1], Microsoft SEAL 3 [p.4], PALISADE 1 [p.4], TensorFlow 1 [p.31]
  • model: SVM 109 [p.1], Logistic regression 4 [p.2], Linear regression 2 [p.12]
  • dataset: Iris 12 [p.14], UCI repository 1 [p.14]
  • hardware: Intel Xeon 2 [p.16], GPU 2 [p.4], FPGA 1 [p.26], CUDA 1 [p.4]
  • technique: Bootstrapping 13 [p.4], Chebyshev 12 [p.4], Sigmoid 4 [p.9], Rescaling 3 [p.7]
  • assumption: RLWE 6 [p.2], Lattice-based 5 [p.1]
  • task: Genomics 6 [p.7], Medical / healthcare 3 [p.9]
  • topic: Homomorphic encryption 51 [p.1], Private / secure inference 14 [p.14], Open source 7 [p.11], Privacy-preserving machine learning 6 [p.14], Benchmark 2 [p.17]

Files

Path (under papers/) Collection Pages Size SHA-256 Note
references/ppml/inference/buchanan-2025_evaluation-of-privacy-aware-svm-learning-using-he-preprint.pdf ppml/inference 31 1.7 MB c4e01af42844 was Buchanan and Ali - 2025 - Evaluation of Privacy-aware Support Vector Machine (SVM) Learning using Homomorphic Encryption.pdf

Database key: evaluation-of-privacy-aware-support-vector-machine-svm-learning-using · created 2026-11-05 · updated 2026-11-05 · edit with python3 tools/paperdb.py set evaluation-of-privacy-aware-support-vector-machine-svm-learning-using COLUMN VALUE