Secured machine learning using Approximate homomorphic scheme for healthcare

Medical disease prediction is a boon after the growth of the medical learning on the recent years The potential for Machine Learning (ML) to boost productivity is significant. To achieve good outcomes, the data quality for training ML models must be great. Any ML algorithm only performs well when it is provided with a tonne of high-quality training data. In order to obtain such high-quality data, several organizations collaborate. It is crucial to uphold data privacy and confidentiality while obtaining data from many patients of different regions. In specific the medical data should be dealt with importance and should be secured to prevent data breach. The proposed work uses CKKS approximate homomorphic encryption scheme to preserve privacy in data when it is fed into the machine learning models. Homomorphic features allow calculations on encrypted data to be done without any privacy leaks, and the user can safely decode the results. This paper proposes a secure mechanism to store and process medical records using homomorphic encryption scheme. The suggested method uses a logistic regression model that was trained using the pima indian diabetes dataset.