Achieving Efficient and Privacy-Preserving Neural Network Training and Prediction in Cloud Environments

The neural network has been widely used to train predictive models for applications such as image processing, disease prediction, and face recognition. To produce more accurate models, powerful third parties (e.g., clouds) are usually employed to collect data from a large number of users, which however may raise concerns about user privacy. In this paper, we propose an Efficient and Privacy-preserving Neural Network scheme, named EPNN, to deal with the privacy issues in cloud-based neural networks. EPNN is designed based on a two-cloud model and techniques of data perturbation and additively homomorphic cryptosystem. This scheme enables two clouds to cooperatively perform neural network training and prediction in a privacy-preserving manner and significantly reduces the computation and communication overhead among participating entities. Through a detailed analysis, we demonstrate the security of EPNN. Extensive experiments based on real-world datasets show EPNN is more efficient than existing schemes in terms of computational costs and communication overhead.

1 Introduction

Recently, the neural network has received considerable attention as it can effectively extract features from high-dimensional data and build predictive models to connect the input data and output results. Due to its high accuracy, neural network has extensive applications in medical prediction[1], image recognition[2], product recommendation [3], etc. The amount of data plays a vital role in successful model training in neural networks, and thus many model requesters (e.g., individuals, companies) seek help from powerful third parties (e.g., clouds) to collect massive data from a large number of users and carry out neural network training. With the trained models, clouds can also provide prediction services, helping users to make better decisions or see their full range of potential options. For example, a healthcare center can collect people's health data and use MLP to train prediction models for different diseases; a company can collect customers’ online activities and use MLP to train models to provide more accurate recommendation services. This “training/prediction as a service” has greatly enhanced data acquiring and processing abilities, thereby opening a new door for neural networks.