Privacy-Preserving Machine Learning for E-Health Applications: A Survey

As more information about people's health is gath-ered and analyzed, privacy concerns have become increasingly essential. This paper summarizes how machine learning (ML) methods are presently used to safeguard privacy in e-health apps. Furthermore, we discuss how privacy-preserving approaches such as differential privacy, secure multi-party computing, and homomorphic cryptography may be utilized to protect privacy in ML. The study also conducts an overview of the literature on how these methods can be used in various e- health scenarios, such as predictive modeling, disease detection, and clinical decision support. Based on the privacy-preserving technique, we classified current privacy-preserving ML work into four categories (i.e., data encryption, data anonymization, model encryption, and model obfuscation). Finally, the article discusses the challenges of applying privacy-preserving machine learning (PPML) in the e-health domain and what types of research could be conducted as open future problems in this field.