Neural Network Training on Encrypted Data with TFHE

TL;DR

This paper introduces a method for training neural networks on encrypted data using fully homomorphic encryption, enabling privacy-preserving collaborative learning on split datasets.

Contribution

It presents a unified training approach that operates on encrypted data and supports collaborative learning with data split across multiple parties.

Findings

Successfully trained logistic regression and MLP models on encrypted data

Supports horizontal and vertical data splits for collaboration

Demonstrates feasibility of privacy-preserving neural network training

Abstract

We present an approach to outsourcing of training neural networks while preserving data confidentiality from malicious parties. We use fully homomorphic encryption to build a unified training approach that works on encrypted data and learns quantized neural network models. The data can be horizontally or vertically split between multiple parties, enabling collaboration on confidential data. We train logistic regression and multi-layer perceptrons on several datasets.