Privacy-Preserving Graph Analytics Engine

图挖掘可以带来巨大的社会和经济价值，但也会泄露个人的敏感信息。虽然差分隐私作为私有数据分析的有力隐私保障已经出现，但现有的差分隐私系统和算法主要支持表格数据。由于图中信息的复杂性和多样性，将现有的表格数据的隐私概念、算法和系统扩展到图数据面临许多挑战。到目前为止，还没有实际的图形数据库系统能够支持对私有图形的分析。本研究的主要目标是建立一个保护隐私的图分析引擎，在保证可证明的隐私保证的同时，准确有效地回答对私有图的交互式查询。我们打算为图形数据形式化一个新颖的、可定制的隐私框架，使用最先进的图形数据库系统支持交互式图形查询，并在分布式设置中启用图形分析。

Graph mining can bring great social and economic values but leak sensitive information about individuals. Though differential privacy has appeared as a strong privacy guarantee for private data analysis, existing systems and algorithms for differential privacy mainly support tabular data. Due to the complex and diverse information within graphs, extending existing privacy notions, algorithms, and systems for tabular data to graph data face many challenges. So far, there are no real-world graph database systems that can support the analysis of the private graphs. The primary objective of this research is to build a privacy-preserving graph analytics engine that can answer interactive queries on private graphs accurately and efficiently while ensuring a provable privacy guarantee. We intend to formalize a novel and customizable privacy framework for graph data, support interactive graph queries using state-of-the-art graph database systems, and enable graph analysis in a distributed setting.