Secret-Shared Database Joins for Non-Unique Match Columns
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing secret sharing techniques are limited to joining database tables with unique matching columns, restricting the types of databases that can be joined, and expose private data when different computers perform a join operation.
Innovation Solution
A method for performing a secret-shared join between a database table with a unique matching column and one with non-unique matching columns by modifying values with identifiers, sorting the combined table, generating a bit vector to identify matches, and combining rows using secure multi-party computation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If secret sharing techniques are used to join database tables, then security of private data is improved, but the types of database tables that can be joined are limited to those with unique matching columns
Solution Approach 1:
The matching column values are segmented by appending identifiers (0 or 1) to indicate which database table they originate from. This segmentation allows the system to distinguish between values from different tables even when the underlying values are not unique, enabling joins on non-unique columns while maintaining security through the identifier-based differentiation mechanism.
Solution Approach 2:
The patent adds a new dimension to the matching column values by appending an identifier bit (0 or 1) to each value. This transforms the original one-dimensional values into two-dimensional identifiers, allowing the system to differentiate between values from different database tables and enabling secure joins on columns that would otherwise be non-unique.
2Reliability
If existing secret sharing join methods are used, then security is maintained, but flexibility in joining different types of databases is reduced
Solution Approach 1:
The patent creates a universal join mechanism that works for both unique and non-unique matching columns by using the identifier appending approach. This multi-functional method allows the same secret sharing protocol to handle various types of database tables regardless of whether their matching columns have unique values, thereby increasing flexibility while maintaining security.
3Reliability
If database tables with non-unique matching columns are joined using traditional secret sharing, then data privacy is protected, but the join operation cannot be performed
Solution Approach 1:
The patent performs a preliminary action by appending identifiers to matching column values before executing the join operation. This preprocessing step transforms non-unique values into unique identifier-composite values, enabling the join operation to proceed successfully while maintaining data privacy through the secret sharing protocol throughout the process.
Data Source
AI summary
Embodiments can perform a database join that allows secret shared database join between a database table with a unique matching column and a database table with a non-unique matching column (i.e., contains unbounded repeats of values).


