Data Analytics over Encrypted Data from Fully Homomorphic Encryption
Sonam Mittal, Soni Singh · 2023
Recently, in machine learning research over private data becomes a need. Over the cloud, open data can be easily stolen and misused. However, dealing with data in every aspect of security and privacy of data as well as secure computation, analysis and prediction with the machine learning and deep learning models is a bit challenging due to several security-threatening issues. This paper helps to incorporate secure computation and analysis using a fully homomorphic encryption model without exposing data. Further, by deploying the Logistic regression, the ML model is used to predict the values over encrypted datasets. The logistic regression helps to predict whether a person is suffering from diabetes or not, securely over an encrypted dataset using the proposed FHE model while maintain the privacy of data without exposing data to any third party. Additionally, the proposed FHE-based ML model is compared with the standard ML model based on different performance measures. The results show that the ML model performs comparably and shows promising results while keeping data private and secure throughout computation and processing of dataset.