A Neoteric Approach to Improvise Privacy Shielding of Sensitive Healthcare Information & Prediction of Disease

Abstract

There has been an ongoing compulsion on health organizations to share data for analysis purposes. The healthcare data includes patient behaviors & records, DNA, laboratory test data, activity log, sensible data, cost data, and demographic data. Privacy becomes supplementally crucial in some scenarios when the data is shared with 3rd party along with the personal information of patients, and confidential record of healthcare organizations. Nonetheless, several suitable guidelines, privacy-preserving laws, and compliance requirements are there to safeguard electroclinic healthcare data. Although, privacy, security breaches and data disclosures remain key issues for healthcare systems. Anonymization techniques are liberating from privacy-related regulations. Moreover, Machine Learning (ML) models can imply anonymized data. Thus, it generates an anonymized secure ML model, which provides greater protection against membership and attribute inference attacks. The heuristic approach results comparatively higher in accuracy where it does not violate data privacy and can be handled to train and test the model with securing high data utility and accuracy. Here we analyze and compare linkage attribute attack possibilities and data loss on the anonymized models. Also mentioned tools available for data anonymization and result of k-anonymity performed using ARX tool while decreasing the risk of attacks. various ML Models are applied to anonymized covid data prediction. Comparative analysis of Machine learning models with train-test accuracy, precision, recall and F1-score. Our security model’s results suggest that the proposed model makes the healthcare data system secure and unauthorized access to protected patient healthcare information almost impossible.