A blockchain-based privacy-preserving asynchronous federated sharing method and system

Abstract

The invention discloses an asynchronous federation sharing method for privacy protection based on blockchain, comprising: S1, formulating contract release information sharing requirements; S2, task participants participating in information sharing; S3, verifying transactions and uploading the blockchain; S4, Perform asynchronous federated learning; S5, verify transactions and upload to the blockchain; and S6, distribute rewards for dynamic incentive sharing. The invention uses federated learning technology to ensure that data sharing between enterprises or departments is only a shared data model instead of local data, and proposes to use the index mechanism in differential privacy to select a local model with a high degree of contribution, and allocate a privacy budget to ensure the security of the local model. It proposes an adjustment mechanism based on a multi-weight model to further improve the utility of the global model. At the same time, blockchain technology is used to ensure the efficiency of transactions and the security of the model. It can make data management, data sharing and collaboration efficient become a reality, while ensuring the security of transactions.

Description

technical field [0001] The invention relates to the technical field of information processing, in particular to a method and system for asynchronous federated sharing of privacy protection based on blockchain. Background technique [0002] In the information age, data has extremely important value, and the importance of data privacy protection has gradually become apparent. Enterprise data is generally multi-source, complex and heterogeneous, including order information, inventory information and logistics information, etc. These data volumes are generally very large and are usually stored on local servers or through central servers. Management, the current enterprise information sharing methods can be divided into traditional information methods and centralized information sharing methods. The traditional information sharing method is that the data management mode is single, and there are data islands between enterprises or departments, which are not connected to each othe...

AI Technical Summary

Problems solved by technology

The traditional information sharing method is a single data management model, there are data islands between enterprises or departments, and they are not connected to each other, and it is difficult to obtain data value by data analysis

The centralized information sharing method manages the data of all parties in a unified manner and stores the data in a central server, but the problem is that the central server is easy to be hacked or the data is stolen due to other reasons, and it is difficult to delegate data usage and management rights , leading to complex data access process; the traditional federated learning model also has this problem, the central server sends the initial model to each participant, and the participant uses the local data set to train the model, which is easy to cause a single point of failure and also exposes privacy risk

[0004] When the inventor implemented the present invention, he found that the existing enterprise information sharing method has the following disadvantages: 1. The data management mode of the traditional enterprise information sharing method is single and not connected to each other, and there are data islands between enterprises or departments

2. In the existing centralized information sharing method of the enterprise, the central server is easy to be hacked or the data is stolen due to other reasons, and there is a single point of failure

3. The existing information sharing methods of enterprises cannot meet the efficient needs of enterprises or departments in information sharing

4. In the synchronous federated learning, the efficiency of model training is low, and the computing resources are idle and wasteful.

Images