A Survey on Privacy-Preserving Learning using Vertical Federated Learning with Representation Synthesis

Abstract

In an era where data is one of the most valuable resourceswehave,protectingithasbecomejustasimportant as using it. Traditional machine learning has long relied on centralizingdatabypullingeverythingintooneplacesoa model can learn from it. However, as datasets become larger and more sensitive, this approach introduces serious risks including privacyviolations,databreaches,andreducedwillingnessamong organizations to share confidential information. ThispaperpresentsaVerticalFederatedLearning(VFL)frame- work designed to allow multiple organizations to collaboratively trainamachinelearningmodelwithoutexchangingrawdata. In this approach, each participating organization holds different featuresofthesameusersandtrainsalocalmodelindependently. Onlyintermediaterepresentationsaresharedwithacentralserver, whichcombinesthemintoaunifiedglobalmodelwhilepreserving privacy.

Introduction

This study proposes a privacy-preserving Vertical Federated Learning (VFL) framework that allows organizations to collaboratively train machine learning models without sharing sensitive raw data. Unlike traditional centralized learning, which raises privacy, security, and regulatory concerns, VFL enables participants to keep data locally while sharing only secure model representations.

The framework allows organizations with different information about the same users to independently preprocess data and train local models. The resulting privacy-preserving embeddings are sent to a central server, where they are combined to build a global model without exposing confidential information.

To strengthen security and trust, the system integrates blockchain technology, encryption, secure aggregation, authentication, and access control mechanisms. Blockchain maintains an immutable record of model updates and participant activities, ensuring transparency, integrity, and tamper resistance.

Conclusion

This paper presented a privacy-preserving learning frame- work using Vertical Federated Learning with Representation Synthesis. The proposed architecture enables multiple organi- zations to collaboratively train machine learning models while ensuring that sensitive raw data remains protected within local environments. By integrating secure aggregation, encryption techniques, and distributed learning mechanisms, the framework enhances privacy, transparency, and collaboration among participating organizations. The system demonstrates that organizationscan achieve high-quality collaborative intelligence without sacrificing confidentiality or regulatory compliance. Future work may focus on reducing communication over- head, improving scalability for large-scale deployments, and integrating advanced privacy-preserving technologies such as fully homomorphic encryption and differential privacy.

References

[1] DilipKumarJangBahadurSainietal.,“PersonalizedFederatedLearningfor Privacy-Preserving and Scalable IoT-Driven Smart Healthcare,” 2025. [2] Ravinder Singh, Smriti Mahajan, Sofia Singh, “Federated Learning:Next-GenPrivacy-PreservingAIFrameworkforConsumerandIndustrialApplications,” 2025. [3] JintaoLiang,SenSu,ZhenyaWang,“VerticalFederatedRepresentationSynthesis for Non-Aligned Samples in the Active Party,” 2025. [4] Zhang et al., “Privacy-Preserving Federated Learning using SecureAggregation,” 2024. [5] Chenetal.,“DifferentialPrivacyinFederatedLearningSystems,”2023. [6] GuttiChandu,ThumulaKarthik,andBalbudheParag(2025)

Copyright

Copyright © 2026 Vyshnavi M, Dhanusha Naik, Bhoomika H, Harshitha B, Jahanavi P Gowda. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.