Ilmu Komputer & AI editorial
Open AccessOA2024
Secure Multi-Party Computation for Machine Learning: A Survey
This survey reviews recent advancements in secure multi-party computation (SMPC) for machine learning, categorizing systems into homomorphic-based and secret-sharing-based approaches, and identifies gaps in security requirements, information exchange, incentives, data authenticity, and operational efficiency.
Ian Zhou; Farzad Tofigh; M. Piccardi; M. Abolhasan; D. Franklin; J. Lipmanยท IEEE Accessยท 2024ยท DOI 10.1109/ACCESS.2024.3388992
The core problem
Machine learning (ML) is a powerful technology for extracting information from data of diverse nature and origin. As its deployment increasingly depends on data from multiple entities, ensuring privacy for these contributors becomes paramount for the integrity and fairness of machine learning endeavors. This review looks into the recent advancements in secure multi-party computation (SMPC) for machine learning, a pivotal technology championing data privacy. The survey evaluates these applications from various aspects, including security models, requirements, system types, and service models, aligning with the IEEEโs recommended practices for SMPC. The motivation stems from the growing need to protect sensitive data while enabling collaborative ML across distributed parties.
Innovation
The authors conduct a literature analysis of SMPC systems for ML, categorizing them broadly into two types: homomorphic-based systems and secret sharing-based systems. Homomorphic-based systems facilitate computations on encrypted data, ensuring data remains confidential throughout the computation. Secret sharing-based systems disseminate data across parties in fragmented shares, such that no single party has access to the complete data. The evaluation framework considers security models, requirements, system types, and service models, in line with IEEEโs recommended practices for SMPC. The survey systematically reviews recent advancements and compares them based on these criteria.
Introduction
Machine learning (ML) is a powerful technology for extracting information from data of diverse nature and origin. As its deployment increasingly depends on data from multiple entities, ensuring privacy for these contributors becomes paramount for the integrity and fairness of machine learning endeavors. This review looks into the recent advancements in secure multi-party computation (SMPC) for machine learning, a pivotal technology championing data privacy. The survey evaluates these applications from various aspects, including security models, requirements, system types, and service models, aligning with the IEEEโs recommended practices for SMPC. The motivation stems from the growing need to protect sensitive data while enabling collaborative ML across distributed parties.
Why it matters
The survey highlights several gaps in the current landscape of SMPC for ML. These include: (1) security requisites โ the need for clearer definitions and standardized security models; (2) streamlined information exchange โ efficient communication protocols to reduce overhead; (3) incentive structures โ mechanisms to encourage participation and honest behavior; (4) data authenticity โ ensuring the integrity and provenance of data; and (5) operational efficiency โ improving scalability and reducing computational costs. Recognizing these challenges, the authors envision a holistic SMPC protocol tailored for machine learning applications that addresses these gaps. Such a protocol would integrate security, efficiency, and incentive considerations to enable practical and privacy-preserving collaborative ML.
Who should read this
Pembaca Research Digest dan praktisi rumpun ini.
Opening member contentโฆ