HE-CloudML: a privacy-preserving framework for secure machine learning inference over encrypted cloud data using homomorphic encryption
- 1Department of Computer Science, Faculty of Computing, SIMAD University Mogadishu
Abstract
The widespread adoption of cloud-based Machine Learning as a Service (MLaaS) exposes sensitive user data to critical privacy risks during inference, as plaintext data must typically be processed by untrusted cloud servers. This paper presents HE-CloudML, a unified privacy-preserving framework for secure deep neural network (DNN) inference over encrypted cloud data using Homomorphic Encryption (HE). HE-CloudML is architected as a three-tier system comprising a client-side CKKS encryption module, a cloud-side HE inference engine, and a key management layer, ensuring that raw input data is never exposed to the server at any stage of computation. The framework introduces HE-compatible polynomial activation function approximations via degree-5 Chebyshev minimax polynomials, an optimized SIMD ciphertext batching strategy exploiting Ring Learning With Errors (RLWE) slot packing, and an adaptive lazy bootstrapping pipeline to substantially reduce homomorphic evaluation depth and inference latency. A formal security analysis under the IND-CPA model grounded in the RLWE hardness assumption shows that the cloud server cannot distinguish, and therefore cannot recover, the plaintext input, intermediate activations, or output from the ciphertexts it observes; this argument covers a semi-honest server and does not by itself establish resistance to active/malicious-server, model-inversion, or membership-inference attacks, which remain future work. Comprehensive experiments across three domains benchmark image classification (MNIST: 99.28%, CIFAR-10: 90.37%), medical imaging (93.61%), and financial fraud detection (96.44%) demonstrate that HE-CloudML achieves near-plaintext accuracy with a maximum accuracy drop of 1.81%, while delivering up to 26.9× latency improvements over CryptoNets.
Related
- Privacy-preserving clustered federated learning via differential privacy and homomorphically encrypted prototypes
- FFT-GA-BC: a fused federated transformer genetic algorithm framework with blockchain-secured, differentially private, homomorphically encrypted aggregation for mutation, phenotype, and binary-allele analysis.
- Cytomegalovirus Retinitis Across the Immunosuppression Spectrum: A Contemporary Framework for Risk Stratification, Diagnosis, and Vision-Preserving Therapy
- Lactate-Informed Exercise Prescription in Mild Cognitive Impairment and Alzheimer's Disease: A Translational Framework for Preserving Cognition, Mobility, and Independence.
- A privacy-preserving multicenter framework for lung nodule segmentation in low-dose CT using an F1-score-weighted federated 3D Transformer-UNet
BibTeX
@article{abdirahman2026hecloudml,
title = {{HE-CloudML: a privacy-preserving framework for secure machine learning inference over encrypted cloud data using homomorphic encryption}},
author = {Abdullahi Ahmed Abdirahman and Abdirahman Osman Hashi and Ubaid Mohamed Dahir and Mohamed Abdirahman Elmi},
journal = {Scientific Reports},
year = {2026},
month = sep,
doi = {10.1038/s41598-026-69466-7},
url = {https://doi.org/10.1038/s41598-026-69466-7},
note = {CC BY},
}
RIS
TY - JOUR
TI - HE-CloudML: a privacy-preserving framework for secure machine learning inference over encrypted cloud data using homomorphic encryption
AU - Abdullahi Ahmed Abdirahman
AU - Abdirahman Osman Hashi
AU - Ubaid Mohamed Dahir
AU - Mohamed Abdirahman Elmi
PY - 2026
DA - 2026/09/01
JO - Scientific Reports
DO - 10.1038/s41598-026-69466-7
UR - https://doi.org/10.1038/s41598-026-69466-7
AB - The widespread adoption of cloud-based Machine Learning as a Service (MLaaS) exposes sensitive user data to critical privacy risks during inference, as plaintext data must typically be processed by untrusted cloud servers. This paper presents HE-CloudML, a unified privacy-preserving framework for secure deep neural network (DNN) inference over encrypted cloud data using Homomorphic Encryption (HE). HE-CloudML is architected as a three-tier system comprising a client-side CKKS encryption module, a cloud-side HE inference engine, and a key management layer, ensuring that raw input data is never exposed to the server at any stage of computation. The framework introduces HE-compatible polynomial activation function approximations via degree-5 Chebyshev minimax polynomials, an optimized SIMD ciphertext batching strategy exploiting Ring Learning With Errors (RLWE) slot packing, and an adaptive lazy bootstrapping pipeline to substantially reduce homomorphic evaluation depth and inference latency. A formal security analysis under the IND-CPA model grounded in the RLWE hardness assumption shows that the cloud server cannot distinguish, and therefore cannot recover, the plaintext input, intermediate activations, or output from the ciphertexts it observes; this argument covers a semi-honest server and does not by itself establish resistance to active/malicious-server, model-inversion, or membership-inference attacks, which remain future work. Comprehensive experiments across three domains benchmark image classification (MNIST: 99.28%, CIFAR-10: 90.37%), medical imaging (93.61%), and financial fraud detection (96.44%) demonstrate that HE-CloudML achieves near-plaintext accuracy with a maximum accuracy drop of 1.81%, while delivering up to 26.9× latency improvements over CryptoNets.
SN - 2045-2322
ER -