A Privacy-Preserving Aggregation Method via Grouped Secure Multi-Party Computation
Ziqing Li, Sheng Hsien Lin, Tianle Li, Pengfei Zhao · 2024
This paper introduces a privacy-preserving method for the collection of aggregate statistics.The proposed system consists of a several groups of servers and multiple clients.Each client holds a private data value and aims to obtain the specific result of all values (including sum, AND/OR, maximum value, etc.), the servers compute the aggregate result of the private values without disclosing any individual data.We divide the clients into several groups, with each group utilizing a distinct set of servers to perform aggregation.The final aggregate statistics are then computed by combining the results from all groups.Our scheme is based on secret-sharing technology in secure multi-party computation (MPC).As long as one server in each group involved in daBits generation is honest, the servers learn almost nothing about the clients' private data.To maintain robustness in the face of malicious clients, we enforce the use of Boolean secret-sharing when clients upload their data, enabling servers to verify whether the data is well-formed.Our evaluation demonstrates that the proposed scheme effectively balances between performance and security.