Privacy Preserving Medical Image Inference Portal Using Homomorphic Encryption

Abstract

The growing reliance on AI for medical image analysis presents important privacy risks when patient data is processed on external servers. To solve this problem, we suggest a Privacy- Preserving Medical Image Inference Portal that uses Homomorphic Encryption (HE) to keep sensitive images encrypted while they are being processed. The system lets users upload encrypted medical images that a neural network processes without decrypting them, which keeps privacy safe. A lightweight model that is optimized for encrypted operations strikes a balance between accuracy and speed of computation. The portal has an easy-to-use interface for safe uploads and result retrieval, showing that advanced cryptography can be used in healthcare di- agnostics. This work shows a safe way to use AI in medical imaging that protects both privacy and usefulness.