A Novel Privacy-Preserving Deep Learning Scheme for the Classification of Covid-19 in Chest X-Ray Images

Due to its efficiency, deep learning is increasingly used in healthcare, necessitating stringent data privacy and protection measures. The DL and privacy communities must work together to overcome the gap. This research presents a privacy-preserving deep learning (PPDL)-based technique for securing Chest X-ray image classification. The suggested framework is based on federated learning and the secure multi-party computation (SMC) protocol with elliptic curve cryptography. Experiment findings on the COVID-19 Radiography database reveal that when choosing a suitable number of clients, our proposed framework achieves an accuracy close to the original centralized setting without privacy enhancement. The proposed model is highly accurate while avoiding privacy leaks.