Privacy Vulnerabilities in Multi-Cloud Machine Learning: Five Integrated Theoretical Frameworks for Systematic Understanding, Assessment, and Mitigation

Fuente: Zenodo
Saved in:
Bibliographic Details
Main Author: Tahchiev, Andrean
Format: Recurso digital
Language:English
Published: Zenodo 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866901816136957952
author Tahchiev, Andrean
author_facet Tahchiev, Andrean
contents <p>Privacy vulnerabilities in cloud-based machine learning systems represent a critical challenge in modern distributed computing environments. This paper presents five integrated theoretical frameworks for systematically understanding and mitigating privacy vulnerabilities in multi-cloud ML environments: (1) a Three-Dimensional Vulnerability Model providing formal mathematical risk categorisation with 87% retrospective classification accuracy; (2) a Protection Mechanism Interaction Framework quantifying synergistic and interfering mechanism combinations; (3) a Deployment-Risk Taxonomy connecting four deployment archetypes to distinct vulnerability profiles covering 92.6% of documented incidents; (4) a multi-objective Protection Selection Model achieving 77% alignment with expert decision analyses; and (5) a standardised Evaluation Metrics Framework. The research employs critical realist methodology, analysing 27 documented privacy incidents. All validation criteria were exceeded across three independent validation phases.</p>
format Recurso digital
id zenodo_https___doi_org_10_5281_zenodo_18979816
institution Zenodo
language eng
publishDate 2026
publisher Zenodo
record_format zenodo
spellingShingle Privacy Vulnerabilities in Multi-Cloud Machine Learning: Five Integrated Theoretical Frameworks for Systematic Understanding, Assessment, and Mitigation
Tahchiev, Andrean
machine learning privacy, cloud computing, multi-cloud, differential privacy, membership inference, vulnerability assessment
<p>Privacy vulnerabilities in cloud-based machine learning systems represent a critical challenge in modern distributed computing environments. This paper presents five integrated theoretical frameworks for systematically understanding and mitigating privacy vulnerabilities in multi-cloud ML environments: (1) a Three-Dimensional Vulnerability Model providing formal mathematical risk categorisation with 87% retrospective classification accuracy; (2) a Protection Mechanism Interaction Framework quantifying synergistic and interfering mechanism combinations; (3) a Deployment-Risk Taxonomy connecting four deployment archetypes to distinct vulnerability profiles covering 92.6% of documented incidents; (4) a multi-objective Protection Selection Model achieving 77% alignment with expert decision analyses; and (5) a standardised Evaluation Metrics Framework. The research employs critical realist methodology, analysing 27 documented privacy incidents. All validation criteria were exceeded across three independent validation phases.</p>
title Privacy Vulnerabilities in Multi-Cloud Machine Learning: Five Integrated Theoretical Frameworks for Systematic Understanding, Assessment, and Mitigation
topic machine learning privacy, cloud computing, multi-cloud, differential privacy, membership inference, vulnerability assessment
url https://doi.org/10.5281/zenodo.18979816