跨信任域的联邦k-支配Skyline查询算法  

Cross trust domain federated k-dominant skyline query processing

作　　者：史烨轩 童咏昕[1,2,3] 周昊 许可[1,2,3] 吕卫锋[1,2] SHI Yexuan;TONG Yongxin;ZHOU Hao;XU Ke;LYU Weifeng(State Key Laboratory of Software Development Environment,Beihang University,Beijing 100191,China;School of Computer Science and Engineering,Beihang University,Beijing 100191,China;Beijing Advanced Innovation Center for Future Blockchain and Privacy Computing,Beihang University,Beijing 100191,China)

机构地区：[1]北京航空航天大学软件开发环境国家重点实验室,北京100191 [2]北京航空航天大学计算机学院,北京100191 [3]北京航空航天大学未来区块链与隐私计算高精尖创新中心,北京100191

出　　处：《大数据》2023年第4期32-43,共12页Big Data Research

基　　金：国家自然科学基金资助项目(No.U21A20516,No.62076017);北航基础研究建设基金资助项目(No.YWF-22-L-531);微众学者计划。

摘　　要：k-支配Skyline查询是一种主流的Skyline查询变种,其在多目标决策与推荐领域有着广泛的应用。随着这些应用规模不断扩大,在由多个参与方组成的数据联邦中进行跨域k-支配Skyline查询的需求日益旺盛。然而,由于数据联邦中的参与方之间彼此不互信,进行跨信任域的查询计算需引入大量安全操作,效率较低。为此提出了一种基于跨域隐私向量聚合的算法,从而实现高效的联邦k-支配Skyline查询,并运用一种密文压缩技术进一步优化查询效率,最后通过充分的实验验证了所提方案的优越性。k-dominant skyline is a prevailing skyline query which has widespread applications in multi-criteria decision making and recommendation.As these applications continuously scale up,there is an increasing demand to support k-dominant skyline over a data federation which consists of multiple data silos,each holding disjoint columns of the entire dataset.Yet it is challenging to support k-dominant skyline over a data federation.This is because strict security constraints are often imposed to query processing over data federations,whereas naively adopting security techniques leads to unacceptably inefficient queries.In this paper,we presented an efficient and secure k-dominant skyline for a data federation.Specifically,we devised a novel private vector aggregation-based solution with ciphertext compression-based optimization for efficient k-dominant skyline query processing while providing security guarantees.Extensive evaluations on both synthetic and real datasets showed the superiority of our method.

关 键 词：k-支配Skyline查询 数据联邦 安全多方计算 同态加密 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]