Secure Multiparty Computation for Privacy‐Preserving Machine Learning in Healthcare: A Comprehensive Survey

ABSTRACT Privacy‐preserving machine learning (PPML) using secure multiparty computation (SMC) is emerging as a promising approach for enabling collaborative data analysis in healthcare while protecting sensitive patient information. This survey provides a comprehensive overview of SMC‐based PPML methods, their applications in healthcare, and the associated challenges. We first introduce the fundamental concepts of SMC and compare it with other privacy‐preserving techniques such as homomorphic encryption (HE) and differential privacy. We then discuss various privacy attacks on machine learning (ML) models, including model extraction, membership inference, and model inversion attacks. The survey examined the key applications of SMC in healthcare machine learning, such as collaborative model training, secure federated learning, privacy‐preserving inference, and secure genome analysis. We highlight the advantages of using SMC for PPML in healthcare, including enhanced data privacy, regulatory compliance, and preservation of data utility. The study also analyzed the major challenges in implementing SMC‐based PPML, including computational overhead, scalability issues, and implementation complexity. Finally, we discuss future research directions, including improving scalability, developing hybrid privacy‐preserving techniques, and addressing regulatory considerations. This survey aimed to provide researchers and practitioners with a comprehensive understanding of the current state and future prospects of SMC‐based PPML in healthcare. This article is categorized under: Algorithms and Computational Methods > Networks and Security Applications of Computational Statistics > Health and Medical Data/Informatics