Hardware Accelerated Privacy-Preserving Ensemble Learning for X-Ray Image Diagnostics

Joseph O’Neill, Lydia Bouzar-benlabiod, Nur Zincir-Heywood
Proceedings of the The 39th Canadian Conference on Artificial Intelligence, PMLR 318:514-525, 2026.

Abstract

The adoption of machine learning (ML) in highly regulated and sensitive domains such as healthcare is constrained by escalating concerns regarding data privacy and stringent legal and regulatory frameworks. Although Privacy-Preserving Machine Learning (PPML) techniques provide strong formal guaranties against data leakage, they frequently incur a non negligible reduction in predictive performance. This inherent privacy–accuracy trade-off constitutes a primary obstacle to the practical deployment of PPML systems. This research introduces a novel PPML framework that leverages hardware acceleration techniques in conjunction with ensemble learning to alleviate accuracy degradation and improve performance simultaneously. X-ray images are ideal for PPML diagnostics, as they provide clear, high contrast visualizations that promote fast detection of complex ailments. The resulting system constitutes a robust PPML architecture that attains state of the art performance, achieving an accuracy of 94% relative to existing single model baselines, while simultaneously reducing false negatives by an average of 62%.

Cite this Paper

BibTeX
@InProceedings{pmlr-v318-o-neill26a, title = {Hardware Accelerated Privacy-Preserving Ensemble Learning for X-Ray Image Diagnostics}, author = {O'Neill, Joseph and Bouzar-benlabiod, Lydia and Zincir-Heywood, Nur}, booktitle = {Proceedings of the The 39th Canadian Conference on Artificial Intelligence}, pages = {514--525}, year = {2026}, editor = {Bouzar-Benlabiod, Lydia and Leung, Carson}, volume = {318}, series = {Proceedings of Machine Learning Research}, month = {25--29 May}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v318/main/assets/o-neill26a/o-neill26a.pdf}, url = {https://proceedings.mlr.press/v318/o-neill26a.html}, abstract = {The adoption of machine learning (ML) in highly regulated and sensitive domains such as healthcare is constrained by escalating concerns regarding data privacy and stringent legal and regulatory frameworks. Although Privacy-Preserving Machine Learning (PPML) techniques provide strong formal guaranties against data leakage, they frequently incur a non negligible reduction in predictive performance. This inherent privacy–accuracy trade-off constitutes a primary obstacle to the practical deployment of PPML systems. This research introduces a novel PPML framework that leverages hardware acceleration techniques in conjunction with ensemble learning to alleviate accuracy degradation and improve performance simultaneously. X-ray images are ideal for PPML diagnostics, as they provide clear, high contrast visualizations that promote fast detection of complex ailments. The resulting system constitutes a robust PPML architecture that attains state of the art performance, achieving an accuracy of 94% relative to existing single model baselines, while simultaneously reducing false negatives by an average of 62%.} }
Endnote
%0 Conference Paper %T Hardware Accelerated Privacy-Preserving Ensemble Learning for X-Ray Image Diagnostics %A Joseph O’Neill %A Lydia Bouzar-benlabiod %A Nur Zincir-Heywood %B Proceedings of the The 39th Canadian Conference on Artificial Intelligence %C Proceedings of Machine Learning Research %D 2026 %E Lydia Bouzar-Benlabiod %E Carson Leung %F pmlr-v318-o-neill26a %I PMLR %P 514--525 %U https://proceedings.mlr.press/v318/o-neill26a.html %V 318 %X The adoption of machine learning (ML) in highly regulated and sensitive domains such as healthcare is constrained by escalating concerns regarding data privacy and stringent legal and regulatory frameworks. Although Privacy-Preserving Machine Learning (PPML) techniques provide strong formal guaranties against data leakage, they frequently incur a non negligible reduction in predictive performance. This inherent privacy–accuracy trade-off constitutes a primary obstacle to the practical deployment of PPML systems. This research introduces a novel PPML framework that leverages hardware acceleration techniques in conjunction with ensemble learning to alleviate accuracy degradation and improve performance simultaneously. X-ray images are ideal for PPML diagnostics, as they provide clear, high contrast visualizations that promote fast detection of complex ailments. The resulting system constitutes a robust PPML architecture that attains state of the art performance, achieving an accuracy of 94% relative to existing single model baselines, while simultaneously reducing false negatives by an average of 62%.
APA
O’Neill, J., Bouzar-benlabiod, L. & Zincir-Heywood, N.. (2026). Hardware Accelerated Privacy-Preserving Ensemble Learning for X-Ray Image Diagnostics. Proceedings of the The 39th Canadian Conference on Artificial Intelligence, in Proceedings of Machine Learning Research 318:514-525 Available from https://proceedings.mlr.press/v318/o-neill26a.html.

Related Material