Data Anonymization via Cross-Dataset Relationship Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data anonymization methods fail to prevent identification of entities when combining anonymized datasets, leading to potential breaches of privacy regulations, especially in big data and Data Lake architectures where multiple datasets are analyzed together.
Innovation Solution
A method that determines relationships between datasets, modifies them to prevent entity identification by masking sensitive information, and provides a metadata-driven approach for proactive and automated data protection, ensuring compliance with governance policies through techniques like primary foreign key relationship discovery and anonymization algorithms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If datasets are combined for analytics, then data usability is improved, but privacy protection deteriorates due to potential entity identification
Solution Approach 1:
The system performs preliminary anonymization processing on datasets before they are combined for analytics. By pre-identifying and masking quasi-identifiers and sensitive attributes in advance, the system enables safe data combination while preventing privacy breaches from occurring during subsequent analytics operations.
Solution Approach 2:
The system introduces an intermediary anonymization layer between the raw datasets and the analytics processing. This intermediary process transforms identifying information into anonymized forms while preserving analytical value, allowing datasets to be combined for analytics without exposing underlying entity identities.
2Object-affected harmful factors
If anonymization is applied to datasets, then privacy protection is improved, but data quality deteriorates due to loss of identifying information
Solution Approach 1:
The system applies different levels and types of anonymization to different attributes within datasets based on their sensitivity and identification potential. Quasi-identifiers receive masking treatment while other attributes retain their original values, preserving data quality for non-sensitive dimensions while protecting privacy for identifying dimensions.
Solution Approach 2:
The system transforms sensitive attributes through parameter changes such as masking, generalization, or aggregation rather than complete removal. This allows the data to retain structural integrity and analytical value while changing the parameter representation of sensitive information to prevent entity identification.
3Object-affected harmful factors
If relationship detection between datasets is performed, then privacy protection is improved by preventing entity identification, but system complexity increases
Solution Approach 1:
The system performs preliminary detection of relationships between datasets and identification of quasi-identifiers before analytics operations occur. By pre-computing relationship metadata and identifying potential identification vectors in advance, the system prevents entity identification without requiring complex real-time analysis during data access.
Solution Approach 2:
The system automatically detects relationships between datasets and identifies anonymization requirements without requiring manual configuration or complex external analysis. The automated relationship detection and quasi-identifier identification processes enable the system to self-determine appropriate anonymization strategies based on the actual data relationships.
Data Source
AI summary
The present disclosure relates to a method for data anonymization of a database system. The method comprises: determining if a first dataset and second dataset of the database system have a relationship indicative of an entity having values in the two datasets. A request may be received from a user for at least one of the first and second datasets. In case the first dataset and second dataset have the relationship, at least one of the first and second datasets may be modified such that the indication of the entity is not accessible to the user. And the requested dataset may be provided.