Efficient and Privacy-Preserving Binary Dot Product via Multi-Party Computation
Fatemeh Jafarian Dehkordi, Elahe Vedadi, Alireza Feizbakhsh, Yasaman Keshtkarjahromi, Hulya Seferoglu
TL;DR
This work tackles the challenge of privacy-preserving binary dot-products in distributed machine learning, focusing on bitwise operations common in tree-based vertical federated learning. It introduces BiMPC, a framework that combines Dot Product via Modular Addition (DoMA) and a Three-Party Oblivious Transfer (triOT) primitive to securely compute $y = \mathbf{a}_1 \odot \cdots \odot \mathbf{a}_N$ under an information-theoretic, semi-honest model. The paper provides a concrete two-party (and an outline for multi-party) BiMPC algorithm, including detailed steps for secure addition, modular operations, and OT-assisted non-linear computations, accompanied by privacy guarantees. The approach is claimed to be efficient and scalable for distributed settings, offering a practical tool for privacy-preserving bitwise operations in VFL and related applications.
Abstract
Striking a balance between protecting data privacy and enabling collaborative computation is a critical challenge for distributed machine learning. While privacy-preserving techniques for federated learning have been extensively developed, methods for scenarios involving bitwise operations, such as tree-based vertical federated learning (VFL), are still underexplored. Traditional mechanisms, including Shamir's secret sharing and multi-party computation (MPC), are not optimized for bitwise operations over binary data, particularly in settings where each participant holds a different part of the binary vector. This paper addresses the limitations of existing methods by proposing a novel binary multi-party computation (BiMPC) framework. The BiMPC mechanism facilitates privacy-preserving bitwise operations, with a particular focus on dot product computations of binary vectors, ensuring the privacy of each individual bit. The core of BiMPC is a novel approach called Dot Product via Modular Addition (DoMA), which uses regular and modular additions for efficient binary dot product calculation. To ensure privacy, BiMPC uses random masking in a higher field for linear computations and a three-party oblivious transfer (triot) protocol for non-linear binary operations. The privacy guarantees of the BiMPC framework are rigorously analyzed, demonstrating its efficiency and scalability in distributed settings.