Privacy-Preserving Deep Learning for Erythemato-Squamous Disease Classification

Erythemato-squamous disease (ESD) is a benign skin condition with a broad spectrum of symptoms, creating diagnostic challenges for physicians. Numerous studies have suggested the use of deep learning for constructing classification models. However, the patient data utilized for model training and inference is highly confidential, and any breach could lead to severe implications. To address this issue, we introduce a deep learning framework that leverages secure multiparty computation to enable secure data sharing among entities while safeguarding privacy. Our framework, anchored on replicated secret sharing, enhances computational speed and curtails communication. We have conducted experiments to verify the performance of our framework, and the results demonstrate its accuracy and efficiency.