Data Anonymization Apparatus for k-Anonymity Compliance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In data processing for anonymization, existing methods face challenges in maintaining analysis accuracy while ensuring a low data loss rate, particularly when applying k-anonymization, where a high number of records need to be deleted or when dealing with a small dataset, leading to degraded analysis precision and detailedness.
Innovation Solution
An information processing apparatus that classifies records into sets based on masking target items using a tree-structured dictionary, calculates the number of records and their ratios, and divides the data to minimize record deletion while maintaining granular anonymization, ensuring k-anonymization compliance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If k-anonymization is applied by deleting records with identification probability greater than 1/k, then anonymization compliance is achieved, but data loss rate increases and analysis accuracy degrades
Solution Approach 1:
The patent divides the data into multiple subsets based on item values, and applies different anonymization strategies to different subsets. Records are segmented into those that can be retained with minor abstraction and those requiring deletion, optimizing the balance between compliance and data preservation
Solution Approach 2:
The patent dynamically adjusts the abstraction level of item values based on the distribution characteristics of the data. By changing parameters such as the granularity of item value abstraction, the system achieves k-anonymization while minimizing data loss and preserving analysis accuracy
2Loss of information
If item values are abstracted to reduce the number of records to be deleted, then data loss rate decreases, but analysis precision and detailedness degrade
Solution Approach 1:
The patent applies different levels of item value abstraction to different items based on their specific characteristics and the distribution of their values. Items that can maintain precision with minimal abstraction are treated differently from those requiring higher abstraction to reduce deletions, optimizing both data retention and analysis precision
Solution Approach 2:
The system dynamically determines the appropriate abstraction level for each item value based on the data distribution and the k-anonymization requirement. This dynamic adjustment allows the system to minimize data loss while preserving as much analysis precision as possible
Data Source
AI summary
An information processing apparatus which anonymizes data composed of records including one or more items through statistical processing, includes a memory and a processor to execute classifying respective records constituting the data into one or more sets based on masking target items indicating items to be masked among the items, a dictionary which expresses categories of item values in a tree structure for each of the masking target items, a selected hierarchy level indicating a hierarchy level selected in the tree structure for each of the masking target items, and the number of records included in the data, and calculating the number of records N of each set and a ratio of records belonging to a set including N records, and dividing the data into one or more pieces of data in a case where the ratio of the records belonging to the set including N records satisfies a predetermined condition.