Anonymous Dataset Generation and Risk Evaluation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for generating anonymous datasets to protect privacy often compromise data authenticity and usability by randomly modifying or distorting data, and risk evaluation of re-identification is inefficient due to repeated computations with original data.
Innovation Solution
A method and device that generate an anonymous dataset by creating an equivalence table and cluster table from critical and quasi-identifier attributes, maintaining original values for anchor attributes, and performing risk evaluation using a partition set and weight table to assess re-identification risk efficiently.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If random modification or data perturbation is used to generate anonymous data, then privacy protection is improved, but data authenticity and usability are reduced
Solution Approach 1:
The patent segments the original dataset into multiple equivalence classes based on quasi-identifier attributes. Each equivalence class contains records that are indistinguishable from one another, providing privacy protection. This segmentation approach maintains data authenticity within each class while protecting individual records, resolving the contradiction between privacy protection and data authenticity.
Solution Approach 2:
The patent introduces an equivalence class as an intermediary structure between the original data and the anonymous data. This intermediary groups records with similar characteristics, allowing privacy protection through aggregation while preserving the statistical properties and usability of the data for analysis purposes.
2Measurement precision
If conventional risk evaluation uses all original data for re-identification assessment, then evaluation completeness is improved, but computational efficiency deteriorates
Solution Approach 1:
The patent extracts a representative subset of records from each equivalence class to perform risk evaluation. Instead of evaluating all original data, it selects sufficient samples that capture the re-identification risk characteristics of the entire equivalence class, maintaining evaluation completeness while dramatically reducing computational requirements.
Solution Approach 2:
The patent applies partial action by evaluating only a portion of the data (representative samples from equivalence classes) rather than the complete dataset. This partial evaluation is sufficient to assess re-identification risk because the equivalence class structure ensures that samples represent the entire class, achieving completeness without full data processing.
Data Source
AI summary
A risk evaluation method and a risk evaluation device for evaluating an anonymous dataset generated according to an original dataset are provided. The risk evaluation method comprises the following steps. Acquiring a plurality of appearing times respectively corresponding to a plurality of original values of the original dataset. Generating a partition set and a weight table according to a sample parameter, an anonymous parameter and the appearing times. Dividing the original dataset into a plurality of data partitions according to the partition set, and generating a penetration dataset according to the weight table and the data partitions, wherein the penetration dataset comprises a plurality of sample data. Comparing each sample data with a plurality of anonymous data of the anonymous dataset to obtain a plurality of matching quantities respectively corresponding to the sample data. And calculating and outputting a risk evaluation result according to the matching quantities.