Adapting Differentially Private Synthetic Data to Relational Databases

Relational databases play a pivotal role in modern information systems and business operations due to their efficiency in managing structured data [39]. According to a Kaggle survey [23], 65.5% of users worked extensively with relational data. Additionally, the majority of leading database management systems (e.g., MySQL and Oracle) are built on relational database principles [35]. These systems organize data into multiple tables, each representing a specific entity, and the relationships between tables delineate the connections between these entities. However, the widespread use of relational databases also carries a significant risk of privacy leakage.