System for decentralized ownership and secure sharing of personalized health data
The system disclosed implements a secure method for facilitating secure exchange of health information among various stakeholders, including data owners or contributors, data requestors or miners, and medical providers, including hospitals, clinics, and research laboratories. Additional aspects of the system provide means for conducting secure research on health data collected from data contributors. Health information is exchanged using a decentralized system that incentivizes data contributors to provide health data to data miners. The data miners, which may be pharmaceutical companies, medical laboratories, or hospitals, use various methods in order to perform research on aggregated contributor data, while maintaining contributor privacy.
1. A computer-implemented method of managing health information, said method comprising:
obtaining health information from a plurality of data contributors; storing the health information in an encrypted format within a memory storage;
utilizing a blockchain system to aid in managing the health information; and
permitting a data miner to access and analyze the health information in the encrypted format while maintaining privacy of the at least one data contributor from the data miner;
creating a federated learning model using model parameters or intermediary analysis statistics with the assistance of a privacy-preserving computing node;
the described process of permitting a data miner to access and analyze the health information in the encrypted format comprising:
the data miner makes encrypted requests to the memory storage to request particular types of data analysis;
the memory storage sequentially sends the encrypted requests to the secure computing nodes, and provides the secure computing nodes with data for secure analysis;
the secure computing nodes encrypt the analysis results and return them to the memory storage;
the memory storage returns the encrypted results to the data miners described.
2. The method of claim 1 , wherein the data contributors do not share raw data in the federated learning model.
3. The method of claim 2 , wherein the federated learning model learns a global statistical model while not sharing the data contributor data.
4. The method of claim 3 , wherein the global statistical model is encrypted.
5. The method of claim 1 , wherein exchange of the model parameters or the intermediary analysis statistics are protected through privacy-preserving computation over encrypted data.
6. The method of claim 1 , wherein the health information is genomic data.
7. The method of claim 6 , wherein the genomic data comprises DNA sequence information or RNA sequence information.
8. The method of claim 1 , wherein the health information is associated with personal information about the plurality of data contributors.
9. The method of claim 8 , wherein the personal information comprises demographic information or socioeconomic information about the plurality of data contributors.
10. The method of claim 1 , wherein the health information has been verified by a third party that has provided a signature, which allows the health information to be included in the blockchain system.
11. The method of claim 1 , wherein the federated learning model utilizes a central node that aggregates local analysis results from a plurality of local nodes and produces an aggregated output, while preserving data security for the local nodes.
12. The method of claim 1 , wherein the federated learning model is created using a federated logistic regression on horizontally or vertically partitioned data.
13. The method of claim 1 , wherein the health information comprises DNA sequence information or RNA sequence information.
14. The method of claim 1 , wherein the health information is associated with personal information about the plurality of data contributors, wherein the personal information comprises demographic information or socioeconomic information about the plurality of data contributors.
15. A computer-implemented system for managing health information, said system comprising:
a memory storage configured to store health information from a plurality of data contributors in an encrypted format;
a blockchain system configured to aid in managing the health information; and
a data miner configured to access and analyze the health information in the encrypted format while maintaining privacy of the at least one data contributor from the data miner;
a federated learning model created using model parameters or intermediary analysis statistics with the assistance of a privacy-preserving computing node;
the data miner configured to access and analyze the health information in the encrypted format comprising:
the data miner makes encrypted requests to the memory storage to request particular types of data analysis;
the memory storage sequentially sends the encrypted requests to the secure computing nodes, and provides the secure computing nodes with data for secure analysis;
the secure computing nodes encrypt the analysis results and return them to the memory storage;
the memory storage returns the encrypted results to the data miners described.
16. The system of claim 15 , wherein the data contributors do not share raw data in the federated learning model.
17. The system of claim 15 , wherein exchange of the model parameters or the intermediary analysis statistics are protected through privacy-preserving computation over encrypted data.
18. Non-transitory computer-readable storage media for managing health information, said non-transitory computer readable media comprising:
computer code for obtaining health information from a plurality of data contributors;
computer code for storing the health information in an encrypted format within a memory storage;
computer code for utilizing a blockchain system to aid in managing the health information; and
computer code for creating a federated learning model using model parameters or intermediary analysis statistics with the assistance of a privacy-preserving computing node;
computer code for permitting a data miner to access and analyze the health information in the encrypted format while maintaining privacy of the at least one data contributor from the data miner, comprising:
the data miner makes encrypted requests to the memory storage to request particular types of data analysis;
the memory storage sequentially sends the encrypted requests to the secure computing nodes, and provides the secure computing nodes with data for secure analysis;
the secure computing nodes encrypt the analysis results and return them to the memory storage;
the memory storage returns the encrypted results to the data miners described.
19. The non-transitory computer-readable media of claim 18 , wherein the data contributors do not share raw data in the federated learning model, and wherein exchange of the model parameters or the intermediary analysis statistics are protected through privacy-preserving computation over encrypted data.
20. A computer-implemented method of managing analysis applications, said method comprising:
providing a plurality of data analysis applications that are configured to access and analyze information in associated with at least one data contributor, wherein the information is stored in an encrypted format within a memory storage and a blockchain system aids in managing the information; and
accepting a selection of at least one data analysis application from the plurality of data analysis applications to access and analyze the information; and
the data analysis applications permit a data miner to access and analyze the health information in the encrypted format while maintaining privacy of the at least one data contributor from the data miner; wherein:
the data miner makes encrypted requests to the memory storage to request particular types of data analysis;
the memory storage sequentially sends the encrypted requests to the secure computing nodes, and provides the secure computing nodes with data for secure analysis;
the secure computing nodes encrypt the analysis results and return them to the memory storage;
the memory storage returns the encrypted results to the data miners described.