Data Anonymization Verification for Privacy and Model Utility

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing data anonymization methods fail to ensure irreversible anonymization, leading to potential privacy breaches and non-compliance with regulations like GDPR and ePrivacy, hindering the use of data for business improvement and model sharing.

Innovation Solution

Implementing a combination of data anonymization techniques such as redacting, embedding, swapping, perturbation, and pseudonymization, followed by irreversible transformation to create anonymized data, and using machine-learning models to test the anonymization process for effectiveness.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If data anonymization techniques are applied, then privacy protection is improved, but data utility for business purposes deteriorates

Engineering Contradiction:

Improveprivacy protectionVSAvoiddata utility

Core Design Contradiction:

ReliabilityVSLoss of information

Solution Approach 1:

The patent applies parameter changes by transforming data through multiple anonymization techniques (redaction, embedding, swapping, perturbation, pseudonymization) that modify data parameters while preserving statistical properties. This allows the data to maintain utility for machine learning models while removing personally identifiable information, thus resolving the contradiction between privacy protection and data utility.

Inventive Principle:

Principle #35Parameter changes

2Ease of operation

If traditional anonymization methods are used, then processing simplicity is improved, but anonymization effectiveness deteriorates

Engineering Contradiction:

Improveprocessing simplicityVSAvoidanonymization effectiveness

Core Design Contradiction:

Ease of operationVSReliability

Solution Approach 1:

The patent merges multiple anonymization techniques (redaction, embedding, swapping, perturbation, pseudonymization) into a comprehensive anonymization system. This combination approach enhances anonymization effectiveness by layering multiple protection mechanisms, making it significantly harder to re-identify individuals compared to single-technique methods, while still maintaining automated processing capability.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The patent implements feedback mechanisms through automated testing procedures that verify anonymization effectiveness. The system tests whether anonymized data can be successfully re-identified and uses these results to validate and improve the anonymization process, ensuring high effectiveness while maintaining operational simplicity through automation.

Inventive Principle:

Principle #23Feedback

3Reliability

If data is anonymized for privacy compliance, then regulatory compliance is improved, but data sharing and model improvement opportunities deteriorate

Engineering Contradiction:

Improveregulatory complianceVSAvoiddata sharing capability

Core Design Contradiction:

ReliabilityVSAdaptability or versatility

Solution Approach 1:

The patent transforms data parameters through anonymization while preserving the statistical and structural properties necessary for machine learning analysis. This allows anonymized data to be shared across organizations for collaborative model training and improvement, maintaining adaptability and versatility for business purposes while ensuring GDPR and ePrivacy compliance through irreversible anonymization.

Inventive Principle:

Principle #35Parameter changes

4Reliability

If multiple anonymization techniques are combined, then anonymization robustness is improved, but system complexity deteriorates

Engineering Contradiction:

Improveanonymization robustnessVSAvoidsystem complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent combines multiple anonymization techniques into a unified system that applies redaction, embedding, swapping, perturbation, and pseudonymization in an integrated manner. This merging approach enhances robustness by creating layered protection that is difficult to penetrate, while the modular architecture of the combined techniques allows for manageable system complexity through standardized processing pipelines.

Inventive Principle:

Principle #5Merging (Combining)

Data Source

PatentUS12591708B2Personal data anonymization with model verification

Publication Date: 2026.03.31 TWILIO INC

AI summary

Anonymization is the process to remove personal information from the data. Once the data is anonymized, the data may be used for creating machine-learning models without the risk of invading anyone's privacy. One of the keys to data anonymization is to make sure that the data is really anonymized so nobody could use the anonymized data to obtain private information. Different techniques for data anonymization are presented. Further, the data anonymization techniques are tested for true anonymization by comparing the results from these techniques to a random method of guessing. If the difference in the results is below a predetermined threshold margin, the data anonymization techniques are safe and ready for use.