Secure and Non-interactive k-NN Classifier Using Symmetric Fully Homomorphic Encryption

Abstract

“Machine learning as a service” (MLaaS) in the cloud accelerates the adoption of machine learning techniques. Nevertheless, the externalization of data on the cloud raises a serious vulnerability issue because it requires disclosing private data to the cloud provider. This paper deals with this problem and brings a solution for the K-nearest neighbors (k-NN) algorithm with a homomorphic encryption scheme (called TFHE) by operating on end-to-end encrypted data while preserving privacy. The proposed solution addresses all stages of k-NN algorithm with fully encrypted data, including the majority vote for the class-label assignment. Unlike existing techniques, our solution does not require intermediate interactions between the server and the client when executing the classification task. Our algorithm has been assessed with quantitative variables and has demonstrated its efficiency on large and relevant real-world data sets while scaling well across different parameters on simulated data.