In-Memory Data Anonymization via Equivalence Classes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current data anonymization techniques are limited in their ability to protect personal information in non-numerical and hierarchical data types, failing to effectively anonymize textual data while maintaining data integrity.
Innovation Solution
The Data Anonymizer (DA) system employs techniques like K-anonymity, L-diversity, and T-closeness to anonymize data by grouping records into equivalence classes based on quasi-identifiers, suppressing explicit identifiers, and maintaining data integrity by selecting attributes with the greatest width for anonymization, using a normalized certainty penalty (NCP) calculation to minimize information loss.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If current anonymization techniques are applied to numerical or hierarchical data, then privacy protection is improved, but the ability to anonymize textual data and maintain data integrity deteriorates
Solution Approach 1:
The patent creates a universal anonymization framework that handles multiple data types (numerical, hierarchical, and textual) through a common approach. The system defines equivalence classes and privacy thresholds that work across different data types, making the anonymization technique adaptable to various data structures while maintaining consistent privacy protection standards.
Solution Approach 2:
The patent changes the parameters of anonymization techniques to accommodate textual data. Instead of relying on numerical ranges or hierarchical structures, the system uses string matching, substring operations, and text-based equivalence class definitions. This parameter adaptation enables the same anonymization framework to effectively process textual data while preserving privacy.
2Reliability
If data is anonymized by grouping into equivalence classes, then privacy protection is improved, but information loss increases
Solution Approach 1:
The patent applies partial anonymization by selectively anonymizing only the attributes necessary to achieve the desired privacy threshold. The system calculates the minimum level of generalization required to meet privacy requirements, avoiding excessive anonymization that would cause unnecessary information loss. This allows the data to retain maximum utility while still providing adequate privacy protection.
Solution Approach 2:
The patent replaces traditional mechanical anonymization methods (such as simple suppression or generalization) with a more sophisticated equivalence class-based approach. By using formal definitions of equivalence classes and privacy thresholds, the system achieves more precise control over the balance between privacy protection and information retention, reducing unnecessary information loss.
3Reliability
If explicit identifiers are suppressed to protect privacy, then privacy protection is improved, but data utility deteriorates
Solution Approach 1:
The patent applies local quality by differentiating between explicit identifiers and quasi-identifiers in the anonymization process. Instead of suppressing all identifying information uniformly, the system selectively handles different types of attributes based on their privacy risk and utility. Quasi-identifiers are generalized or suppressed only to the extent necessary to achieve privacy thresholds, while maintaining data utility for analysis purposes.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Disclosed herein are system, method, and computer program product embodiments for data anonymization in an in-memory database. An embodiment operates by identifying a textual attribute corresponding to data of an input table. A value corresponding to the textual attribute is determined for each of the plurality of records. A plurality of groups is generated based on the determined values. At least portion of the data as sorted into the plurality of groups is provided.