Fully Homomorphic Encryption-based Machine Learning for Secure Multi-class Tumor Classification

Abstract

ML outsourcing is one approach to building AI solutions but involves sharing a client’s data with external parties and thus, poses risks on data privacy. Genomic data holds properties that distinguish it from traditional medical data and make it more sensitive with regards to data confidentiality. As patients’ information is at risk with ML outsourcing, there is a need to build a solution that would address this problem. That is, a system that would allow for ML outsourcing in the medical field that would ensure data confidentiality. This paper proposes fully homomorphic encryption-based machine learning to achieve a secure multi-class tumor classifier using ConcreteML.