满足差分隐私的多方水平划分数据合成机制

Multi-party horizontally partitioned data synthesis mechanism with differential privacy

隐私保护的多方数据发布问题因其应用前景广泛而备受关注。针对隐私保护的多方数据发布问题,提出满足差分隐私的多方水平划分数据合成机制DP-MHDS。通过生成低维边缘分布实现对高维数据的降维,利用分布式拉普拉斯机制有效保护了多方聚合边缘分布的隐私。实验结果证实了DPMHDS能够生成高效用的多方合成数据集。

The problem of multi-party data and publication under privacy protection has attracted much attention because of its wide application prospects.To solve the problem of multi-party data publication while ensuring data privacy,a Multi-party Horizontally Partitioned Data Synthesis mechanism based on Differential Privacy(DP-MHDS)is proposed.It reduces the dimension of highdimensional data by generating low-degree marginal distributions and uses distributed Laplace perturbation mechanism to effectively protect the privacy of marginal distributions that aggregated from those parties.DP-MHDS can obtain a multi-party synthesized dataset with high utility is confirmed by experiments.