Tools
Tools
Pulido-Gaytan, Bernardo, González-Vélez, Horacio and Tchernykh, Andrei (2026) Cryptographically Computable Self-Learning Activation Functions for Privacy-Preserving Logistic Regression with Homomorphic Encryption. In: 2026 IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW). IEEE, New Orleans, LA, USA, pp. 571-579. ISBN 979-8-3195-3090-5
Preview |
PDF
Download (600kB) | Preview |
Abstract
Efficient processing of sensitive data in cloud environments requires not only accurate analysis but also rigorous privacy guarantees. The new generation of cloud-based Machine Learning (ML) services therefore demands privacy-preserving techniques that allow robust data analysis and individual privacy to effectively coexist. Homomorphic Encryption (HE) addresses privacy concerns by enabling computations over encrypted data, providing security and confidentiality to users even when execution takes place on untrusted shared infrastructures. Lattice-based HE schemes rely on the hardness of the Ring Learning with Errors problem and natively support only addition and multiplication operations, which poses significant challenges for efficiently implementing non-linear models. In this paper, we design and experimentally evaluate a specialized approach based on cryptographically computable self-learning polynomials to build accurate, low-complexity, and privacy-preserving Logistic Regression (LR) models for confidential data processing. We propose a method to increase the accuracy and performance of privacy-preserving LR with HE (LR-HE) using Self-Learning Activation Functions (SLAFs). SLAFs are polynomials with trainable coefficients that are updated during training, together with synaptic weights, to learn task-specific and LR-specific features. Our results demonstrate the feasibility of applying trained, problem-specific polynomial activations to generate HE-compliant logistic functions, enabling efficient, accurate, and privacy-preserving LR solutions in privacy-sensitive cloud settings.
| Item Type: | Book Section |
|---|---|
| Additional Information: | © 2026 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. |
| Uncontrolled Keywords: | Homomorphic encryption; logistic regression; polynomial approximation; privacy-preserving; self-learning activation functions |
| Subjects: | Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Cloud computing B Philosophy. Psychology. Religion > BJ Ethics > Conduct of life > Reliability > Information integrity > Data integrity Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning |
| Divisions: | School of Computing > Staff Research and Publications |
| Depositing User: | Tamara Malone |
| Date Deposited: | 24 Aug 2026 13:43 |
| Last Modified: | 24 Aug 2026 13:43 |
| URI: | https://norma.ncirl.ie/id/eprint/9610 |
Actions (login required)
 |
View Item |