Support vector machine classification over encrypted data

doi:10.1007/s10489-021-02727-2

Return

Support vector machine classification over encrypted data

2021-08-25

PRE

DOI：10.1007/s10489-021-02727-2

Original

Original request for help

Save

Abstract

En 中文

Support vector machine is one of the most extensively used machine learning algorithms. A typical application scenario of support vector machine classification is the client-server model where a client holding an image requests a server holding a support vector machine model to provide image classification services. However, applying support vector machine in the client-server model requires careful attention to maintain data privacy. In this paper, we will propose a privacy-preserving support vector machine classification scheme based on fully homomorphic encryption. Our scheme guarantees that the server cannot learn the user's image while providing classification service and the client eventually obtains the classification result without learning about the support vector machine model. We introduce several novel techniques to optimize our proposal. Specifically, we use the single instruction multiple data techniques to accelerate the linear component and the approximate method to compute the non-linear component. Our scheme significantly improves computational and communication performance compared to the state-of-the-art work. Experimental results show that our scheme takes only seconds to classify an encrypted image while the state-of-the-art work takes several hundred seconds.

Keywords:

Privacy-preserving computation

Support vector machine

Cloud computing security

Fully homomorphic encryption

Computation over encrypted data

AI Summary

Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.