Privacy-Preserving COVID-19 Screening Using Homomorphic Encryption and Logistic Regression

As COVID-19 expands globally, dependable health screening systems protecting people's privacy are more important than ever. Traditional screening techniques demand the provision of sensitive health data, which raises reasonable privacy concerns. This paper utilizes Homomorphic Encryption (HE) and Logistic Regression to propose a novel architecture for reliable COVID-19 screening, thereby safeguarding patient data. Using a partially homomorphic encryption method, the proposed methodology encrypts patient characteristics, including symptoms, travel history, and exposure risk, allowing for direct logistic regression on encrypted data. The model is trained and evaluated using synthetic datasets imitating real-world screening scenarios. Experimental results show that, with almost little extra processing effort, the encrypted logistic regression model beats its plaintext counterpart (up to 93.5% accuracy). Other significant system features include meeting strict data security requirements and guaranteeing end-to-end data confidentiality. These results suggest that screening tools provided by HE might help to create a balance between data value and privacy, which is a significant consideration in developing pandemic response strategies.

Перевод пока недоступен