Streaming Data Anonymization via Region Partitioning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data anonymization techniques face challenges in maintaining privacy and compliance with legal requirements when dealing with streaming data, particularly in ensuring k-anonymity and protecting sensitive information like location data, especially when dealing with limited memory and real-time data processing.
Innovation Solution
The method involves partitioning a space into regions based on historical data to achieve k-anonymity by assigning data elements to regions and recording sensitive information when the buffer is full, using techniques like PickMax and QuadTree decomposition to ensure that each recorded region contains at least k data points, thereby anonymizing location data effectively.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data elements are assigned to regions and recorded when buffer is full to achieve k-anonymity, then privacy protection is improved, but data analysis accuracy deteriorates due to loss of precise location information
Solution Approach 1:
The space is partitioned into multiple regions, and data elements are assigned to specific regions based on their location. This segmentation allows the system to protect privacy by generalizing exact locations to region-level information while maintaining the ability to perform spatial analysis at the region level, thus resolving the contradiction between privacy protection and data analysis accuracy.
2Reliability
If the buffer size is increased to store more data elements before recording, then k-anonymity is improved, but memory usage deteriorates
Solution Approach 1:
Multiple data elements are merged into region-based aggregates before being recorded to non-volatile storage. By combining data at the region level rather than storing individual data elements, the system achieves k-anonymity requirements while significantly reducing the memory burden, as the buffer only needs to track region counts rather than individual records.
3Reliability
If the recorded region size is reduced to improve privacy, then privacy protection is improved, but data utility deteriorates
Solution Approach 1:
The system applies different levels of generalization to different regions based on their characteristics and sensitivity. High-traffic or sensitive regions may use coarser granularity for better privacy protection, while less sensitive regions can maintain finer resolution to preserve data utility. This local quality approach allows the system to optimize the privacy-utility tradeoff for each region individually.
Data Source
AI summary
Techniques are provided for anonymizing data in a data stream. In various embodiments, data in a data stream is anonymized by obtaining a partitioning of a space into a plurality of regions based on previously obtained data elements associated with the space; receiving a data stream comprising a plurality of newly obtained data elements associated with the space, each comprising an identifying portion and an associated sensitive portion; assigning each of the plurality of newly obtained data elements to one of the plurality of regions based on the identifying portion; determining if a full buffer criterion is satisfied for a buffer that stores at least a portion of the plurality of newly obtained data elements; and selecting one of the plurality of regions having a largest number of the newly obtained data elements and recording the associated sensitive portions of the newly obtained data elements in the selected one of the plurality of regions to a non-volatile storage medium on the condition that the full buffer criterion is satisfied.