PBM-VFL: Vertical Federated Learning with Feature and Sample Privacy

TL;DR

This paper introduces PBM-VFL, a communication-efficient vertical federated learning algorithm that guarantees differential privacy for features and samples by combining secure computation with a novel Poisson Binomial Mechanism.

Contribution

It proposes a new VFL algorithm that integrates differential privacy with secure computation, defining feature privacy and analyzing privacy-utility trade-offs.

Findings

PBM-VFL achieves high privacy guarantees with competitive accuracy.

Theoretical analysis links privacy budget, convergence, and communication costs.

Empirical results demonstrate effective privacy preservation with minimal performance loss.

Abstract

We present Poisson Binomial Mechanism Vertical Federated Learning (PBM-VFL), a communication-efficient Vertical Federated Learning algorithm with Differential Privacy guarantees. PBM-VFL combines Secure Multi-Party Computation with the recently introduced Poisson Binomial Mechanism to protect parties' private datasets during model training. We define the novel concept of feature privacy and analyze end-to-end feature and sample privacy of our algorithm. We compare sample privacy loss in VFL with privacy loss in HFL. We also provide the first theoretical characterization of the relationship between privacy budget, convergence error, and communication cost in differentially-private VFL. Finally, we empirically show that our model performs well with high levels of privacy.