Multi-party Security Computation with Differential Privacy over Outsourced Data

摘要

Differential privacy has received considerable attention for privacy-preserving machine learning applications. In particular, in the cloud computing environment, data are outsourced from different users. Processing outsourced computations on the joint distribution of multiparty's data under multiple public keys with differential privacy is a significant and difficult problem. In this paper, we propose a scheme named 1, multi-party security computation with differential privacy over outsourced data (MSCD) by using a combination of public-key encryption with a double decryption algorithm (DD-PKE) and e-differential privacy to solve this problem. In our work, the cloud server adds the corresponding different statistical noises according to different queries of the data analyst, which differs from previous works in which noise is added by the data provider. In the random oracle model, our scheme is proven to achieve the goal of outsourced computation on the data sets of multiple parties without privacy leakage.