Data Anonymization Risk Assessment for Re-Identification Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing anonymization methods, such as pseudonymization and synthetic data, are inadequate for ensuring data privacy compliance due to reversibility and reliability issues, and current risk assessment tools like WP29 criteria are too strict, leading to excessive data degradation or inefficiency.
Innovation Solution
A data anonymization system that assesses individualization, correlation, and inference risks using a risk assessment tool, allowing selective transformation of high-risk data items to minimize degradation while maintaining data quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If pseudonymization is used to anonymize data, then the risk of individual identification is reduced, but the data remains reversible and can be easily reconstructed to personal data
Solution Approach 1:
The patent replaces traditional mechanical pseudonymization methods with a synthetic data generation approach using generative adversarial networks (GANs). Instead of simply masking or replacing identifiers, the system creates entirely new synthetic datasets that replicate the statistical properties and relationships of original data without containing any actual personal information, thereby eliminating reversibility while preserving data utility
Solution Approach 2:
The system transforms the fundamental parameters of data representation by using GANs to generate synthetic data with matching statistical distributions. The generator network learns the underlying data distribution and produces synthetic records that indistinguishably mimic real data patterns, while the discriminator network ensures the synthetic data meets quality standards, achieving anonymization that cannot be reversed to original personal data
2Measurement precision
If WP29 criteria are strictly applied to assess anonymization risk, then individualization, correlation, and inference risks are evaluated, but the anonymization becomes overly rigorous and significantly degrades data quality
Solution Approach 1:
The patent applies partial action by selectively focusing risk assessment on the three critical dimensions identified by WP29 (individualization, correlation, inference) rather than applying all possible anonymization measures. The GAN-based synthetic data generation is designed to specifically address these three risk types while maintaining data quality, avoiding excessive anonymization that would degrade utility
Solution Approach 2:
The system implements feedback mechanisms through the discriminator network in the GAN framework, which continuously evaluates generated synthetic data against quality standards and statistical properties of original data. This feedback loop ensures that anonymization measures effectively address WP29 criteria while preserving data quality, allowing iterative refinement until both privacy protection and data utility requirements are met
3Object-affected harmful factors
If synthetic data is generated to protect individuals, then real personal information is not contained, but the data generation process is time-consuming and requires creating a new model for each dataset
Solution Approach 1:
The patent enhances productivity by designing the GAN framework with transfer learning capabilities and pre-trained components that can be adapted to different datasets without requiring complete retraining. The generator and discriminator networks can be fine-tuned on new data with significantly reduced training time compared to creating entirely new models, making synthetic data generation more efficient and scalable across multiple datasets and use cases
4Reliability
If traditional anonymization methods are applied to ensure compliance, then data privacy is protected, but the data utility and quality are significantly reduced
Solution Approach 1:
The patent replaces traditional mechanical anonymization techniques (masking, generalization, suppression) with a machine learning-based synthetic data generation system. This substitution allows the preservation of complex relationships, patterns, and statistical properties in the data while ensuring complete anonymization through synthetic generation, thereby maintaining data utility for research and analysis purposes without compromising privacy compliance
Data Source
AI summary
The invention relates to a program based on a data anonymization system wherein the system comprises a module for identifying data and assigning an exposure level with respect to individuals inside or outside the user's organization; a module for identifying feared events and their severity; a module for assessing a legal anonymization criteria avoiding re-identification of individuals; a module for evaluating the level of exploitability of the data; a module for assessing the overall risk of the dataset, in which data with a significant level of exploitability and/or severity are identified; and a module for correcting data and implementing countermeasures or transforming data with a significant level of exploitability and/or severity, in order to reduce that level.