Federated Learning Data Management via Usage Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Data providers face challenges in managing privacy data, particularly in federated learning, as they struggle to track and control the usage of their data, leading to potential decreases in data provision for personalized medical care, due to unclear usage histories and difficulties in opting out.
Innovation Solution
A storage apparatus that stores access policies for each entity, including operation attributes, and maintains operation and data logs to track usage, allowing for transparent data management by specifying usage conditions and maintaining data access authorities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If privacy data is used for federated learning to provide personalized medical care, then the quality of medical care is improved, but data providers face difficulty in tracking and controlling data usage
Solution Approach 1:
The system implements a feedback mechanism by storing operation logs and data logs that record every usage of privacy data. When data is accessed or processed, the system automatically logs the operation details (including which data was accessed, by whom, and for what purpose) and stores this information in the storage apparatus. This feedback loop enables continuous monitoring and tracking of data usage throughout the federated learning process.
Solution Approach 2:
The invention introduces an intermediary storage apparatus that acts as a mediator between data providers and the federated learning system. This storage apparatus maintains separate entity lists for data providers and stores operation logs and data logs as intermediaries that track usage without requiring direct communication or control from data providers during the learning process.
2Measurement precision
If comprehensive data is collected for fine-grained patient segmentation, then the precision of personalized care is improved, but data providers become more concerned about data security and control
Solution Approach 1:
The system segments the tracking of data usage by creating separate entity lists for different data providers and storing operation logs and data logs in a centralized storage apparatus. This segmentation allows fine-grained tracking of individual data provider usage patterns while maintaining overall system functionality. Each data provider's data usage can be independently monitored and controlled.
Solution Approach 2:
The storage apparatus provides feedback to data providers about how their data is being used by storing and making accessible operation logs and data logs. This feedback mechanism addresses data provider concerns by transparently showing usage patterns, enabling them to make informed decisions about data sharing while maintaining the precision needed for personalized care.
3Reliability
If data usage is not transparently tracked, then data access authority is maintained, but opt-out application becomes difficult for data providers
Solution Approach 1:
The system stores operation logs and data logs that provide transparent feedback to data providers about their data usage. When a data provider wishes to opt-out, they can review the stored logs to understand exactly how their data has been used, making the opt-out process more informed and manageable while maintaining their data access authority.
Solution Approach 2:
The storage apparatus serves as an intermediary that maintains data access authority while simultaneously providing transparent tracking. It stores entity lists and logs as intermediaries that can be queried by data providers to understand usage patterns, enabling informed opt-out decisions without compromising the ongoing data access rights of other users.
4Difficulty of detecting and measuring
If detailed operation and data logs are stored for each entity, then data usage tracking is improved, but system complexity increases
Solution Approach 1:
The system merges the tracking of multiple data providers into a unified storage apparatus that maintains entity lists and operation logs centrally. By combining the management of multiple data providers' data into a single integrated system with standardized log storage structures, the overall system complexity is reduced while maintaining detailed tracking capabilities for each individual provider.
Data Source
AI summary
For each entity, an access policy is provided including an access authority for each n-th order data for use in an application or a model. For each operation, provided are an operation log, and a data log that is a log of data of a source or a target of the operation and is associated with the operation log. Provided is an entity list based on an access policy for an operation log and/or a data log. In response to a request, when one or more entity lists in which an entity specified on the basis of the request is recorded are found from a plurality of entity lists, the processor specifies a usage condition on the basis of one or more operation logs and one or more data logs specified using the one or more entity lists, and returns data indicating the specified usage condition to a request source.