Privacy-preserving knowledge graph merging

• Main Author: Rajendran, Reenashini
• Other Authors: Sourav Sen Gupta
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2022
• Subjects: Engineering::Computer science and engineering::Data::Data encryption
• Online Access: https://hdl.handle.net/10356/156604
• Institution: Nanyang Technological University

Knowledge Graphs (KG) empower the creation of intelligent systems that can integrate large amounts of data to generate meaningful insights that remain hidden in traditional databases (e.g. Relational databases). Recognizing the robustness of knowledge graphs, big tech companies have spearheaded the adoption of knowledge graphs for data analytics and management, search engines, intelligent agents, and many other applications. In recent years, an increasing number of organisations from all stripes of industries are trying to adopt knowledge graphs to gain a competitive edge. However, the benefits of KGs are limited by data silos within organisations. Organisations often segregate their data into silos due to their security policies. This hinders the formation of a unified KG that is rich and meaningful to the organisation. This project focuses on utilizing Private Set Intersection (PSI), a secure multi-party computation cryptographic technique, to perform Privacy-Preserving KG Merging on isolated data sources. This paper first explores and outlines key concepts and research relevant to Privacy-Preserving KG Merging and PSI. Then, we present SecureKGMerge, a system that performs Privacy-Preserving KG Merging using PSI. SecureKGMerge is designed with the goal of merging isolated data sources while safeguarding their confidentiality and generating meaningful insights from the KG merge. SecureKGMerge was implemented and tested in a simulated bank consisting of isolated departments. SecureKGMerge was used to detect possible money-laundering activities hidden within the data of these isolated departments, supporting the bank in its anti-money laundering efforts. Tests demonstrated that SecureKGMerge performs accurately and predictably in accordance with its design requirements. Therefore, proving the sufficiency of PSI to perform Privacy-Preserving KG Merging on isolated data sources.