Pneumonia Risk Assessment on Data Encrypted using a Homomorphic Cryptosystem

Cloud system offers several advantages compared to the use of local server. However, for data that requires high level of security, including medical records, privacy issue becomes a problem, especially when the data needs to be analyzed/manipulated. This includes the analysis of patient's medical record to predict Pneumonia risk level. In this paper, a model to solve the problem is proposed. The patient's medical record is first encrypted using Paillier cryptosystem before it is uploaded to the cloud. Then, the record is analyzed, while still being encrypted, to predict the patient's Pneumonia risk level. Modification on the rules of Pneumonia Severity Index (PSI) must be conducted to allow the rules to operate on encrypted data as their input. The evaluation shows that the proposed model can classify all classes, i.e. class I II, III, IV and V, correctly with exactly the same PSI score, reaching the accuracy of 100%.