PE-HEALTH: Enabling Fully Encrypted CNN for Health Monitor with Optimized Communication

摘要: Cloud-based Convolutional neural network (CNN) is a powerful tool for the healthcare center to provide health condition monitor service. Although new service has future prospects in medical, patient's privacy concerns arise because of sensitivity medical data. Prior works address concern have following unresolved problems: 1) focus on data but neglect protect machine learning model itself; 2) introduce considerable communication costs CNN inference, which lowers quality cloud server. To push forward this area, we propose PE-HEALTH, privacy-preserving framework that supports fully-encrypted (both input and model). In Internet Things (IoT) sensor serves as collector. For protecting patient privacy, IoT additively shares collected uploads shared server, efficient suited energy-limited sensor. keep PE-HEALTH allows previously deploy, then, use an encrypted During inference process, does not need servers exchange any extra messages operating convolutional operation, can greatly reduce cost.