Private processing of database queries protects the confidentiality of sensitive data when queries are answered. It is important to design collusion-resistant protocols ensuring that privacy remains protected even when a certain number of honest-but-curious participants collude to share their knowledge in order to gain unauthorised access to sensitive information. A novel setting arises when aggregated queries need to be answered for a large distributed database, but legal requirements or commercial interests forbid making access to records in each subdatabase available to other counterparts. For example, a very large number of medical records may be stored in a distributed database, which is a union of several separate databases from different hospitals, or even from different countries. The present article introduces and investigates two protocols for collusion-resistant private processing of aggregated queries in this novel setting: Accelerated Multi-round Iterative Protocol (AMIP) and Restricted Multi-round Iterative Protocol (RMIP). We define a large collection of query functions and show that AMIP and RMIP protocols can answer all queries in this collection. Our experiments demonstrate that the AMIP protocol outperforms all other applicable algorithms, and this achievement is especially significant in terms of the communication complexity.

中文翻译：

---

用于分布式数据库中聚合查询的私有处理的抗共谋协议

数据库查询的私有处理可在回答查询时保护敏感数据的机密性。设计防串通协议很重要，即使一定数量的诚实但好奇的参与者串通共享他们的知识以未经授权访问敏感信息，也能确保隐私得到保护。当需要为大型分布式数据库回答聚合查询时，会出现一种新的设置，但法律要求或商业利益禁止其他对应方可以访问每个子数据库中的记录。例如，大量的病历可能存储在分布式数据库中，该数据库是来自不同医院甚至来自不同国家的多个独立数据库的联合。本文介绍并研究了两种协议，用于在这种新颖的环境中对聚合查询进行抗合谋私有处理：加速多轮迭代协议 (AMIP) 和受限多轮迭代协议 (RMIP)。我们定义了大量的查询函数集合，并表明 AMIP 和 RMIP 协议可以回答这个集合中的所有查询。我们的实验表明 AMIP 协议优于所有其他适用的算法，这一成就在通信复杂性方面尤为重要。我们定义了大量的查询函数集合，并表明 AMIP 和 RMIP 协议可以回答这个集合中的所有查询。我们的实验表明 AMIP 协议优于所有其他适用的算法，这一成就在通信复杂性方面尤为重要。我们定义了大量的查询函数集合，并表明 AMIP 和 RMIP 协议可以回答这个集合中的所有查询。我们的实验表明 AMIP 协议优于所有其他适用的算法，这一成就在通信复杂性方面尤为重要。