Financial data sharing system and method based on block chain and federal learning

Abstract

The invention relates to the technical field of financial data sharing, and discloses a financial data sharing system and method based on a block chain and federated learning, and the system comprises a block chain layer which records an operation log, a data access record and authority management information in a data sharing process through a distributed ledger technology, executes a data sharing protocol through a smart contract, and stores the operation log, the data access record and the authority management information in the data sharing process; a credible cooperation basis is provided for the federated learning layer; and a multi-chain architecture module in the block chain layer is used for supporting multi-source data interaction and realizing consistency of data sharing and intelligent contract execution among different block chain networks through a cross-chain communication protocol. The block chain technology and the federated learning technology are adopted, and distributed model training and data sharing are carried out on the premise that original data are not shared. The block chain layer records all data operation logs and authority management information, and ensures the security and compliance of data access through an intelligent contract.

Description

Technical Field

[0001] The present invention relates to the technical field of financial data sharing, and specifically to a financial data sharing system and method based on blockchain and federated learning. Background Art

[0002] In the modern financial industry, data has become an extremely important asset. Financial institutions accumulate high-value data through their operations. The data covers customer credit scores, transaction records, risk assessments, market forecasts and other fields, and has extremely high commercial value. With the increase in data volume, how to effectively use data to improve business decisions and optimize risk management has become one of the major challenges facing financial institutions. Existing data sharing technologies have significant deficiencies in privacy protection, security, trust building and fair incentives, which restricts data sharing and collaboration among financial institutions.

[0003] Traditional data sharing methods rely on centralized systems or third-party intermediaries to exchange data, which can easily lead to data privacy leaks and security issues. In the financial field, due to the high sensitivity of data, once a data leak occurs, it will cause serious economic losses and legal consequences. Trust issues between financial institutions also further hinder data sharing. The lack of trust between institutions to directly share data, fears that data will be abused or unauthorized secondary use, affects the efficiency of collaboration between institutions. The existing incentive mechanism cannot fairly evaluate the contributions of each participant, resulting in an imbalance in incentive distribution, which in turn affects the enthusiasm of participants and their willingness to share data.

[0004] The introduction of blockchain technology provides new ideas for data sharing, but blockchain applications are mainly concentrated in a single blockchain network and lack support for cross-chain data interaction. In real applications, different financial institutions use different blockchain networks. How to achieve secure data transmission and cross-chain execution of smart contracts between different blockchain networks has become a technical problem that needs to be solved urgently.

[0005] To this end, those skilled in the art provide a financial data sharing system and method based on blockchain and federated learning to solve the problems raised in the above background technology. Summary of the invention

[0006] In view of the shortcomings of the existing technology, the present invention provides a financial data sharing system and method based on blockchain and federated learning to solve the problems existing in the existing technology in privacy protection, security, trust building and fair incentives.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions: a financial data sharing system based on blockchain and federated learning, comprising: The blockchain layer uses distributed ledger technology to record operation logs, data access records, and permission management information during the data sharing process, and uses smart contracts to execute data sharing agreements, thereby providing a trusted collaboration foundation for the federated learning layer. The multi-chain architecture module in the blockchain layer is used to support multi-source data interaction and achieve consistency in data sharing and smart contract execution between different blockchain networks through cross-chain communication protocols. The Merkle proof mechanism calculation formula used in cross-chain communication is: ; in, Indicates that a hash operation is performed on the data. Indicates string concatenation; The federated learning layer, relying on the security of the blockchain layer, realizes distributed model training and parameter sharing among multiple financial institutions, and ensures data privacy through encryption technology and secure computing methods; The model training module of the federated learning layer includes a weighted average algorithm based on multi-party participation, which is used to perform model training and parameter updates without sharing the original data. The calculation formula of the weighted average algorithm is: ; in, is the weight of the global model, For the The model weights of the participants, For the The amount of data for each participant; The data management layer manages data encryption, access control, and privacy protection based on the needs of the blockchain layer and the federated learning layer to provide data security support; The distributed key management module in the data management layer generates and distributes encryption keys using the Shamir secret sharing algorithm to ensure that k key fragments are required during the decryption process. The key generation formula is: ; in, is the generated key fragment polynomial, is the secret value, is a randomly generated coefficient; The user interface layer serves as an interactive interface for data sharing and collaborative analysis among financial institutions. It combines blockchain and federated learning functions to achieve traceability of operations and transparency of collaboration. The incentive and equity distribution layer records the contribution evaluation results through blockchain and distributes equity and manages incentives based on them, ensuring fairness and traceability in the data sharing process.

[0008] Preferably, the blockchain layer includes: Multi-chain architecture module to support the integration and collaboration of multiple blockchain networks; The cross-chain communication module is used to achieve secure transmission of data and status between different blockchain networks, and ensure data consistency through cross-chain communication protocols.

[0009] Preferably, the federated learning layer includes: Model training module, which is used to perform distributed model training and model updates without sharing the original data; Parameter aggregation module, used to collect and aggregate model parameters uploaded by various participants; The parameter aggregation module combines differential privacy technology to protect the data privacy of each participant by introducing noise in the parameter update process. The noise introduction formula is: ; in, are the updated model parameters, is the learning rate, is the gradient of the loss function, is Gaussian noise.

[0010] Preferably, the parameter aggregation module uses a homomorphic encryption algorithm to transmit encrypted model parameters between participants to ensure privacy and security during parameter transmission. The homomorphic encryption formula is: ; in, is the model parameter The encryption result is is the public key exponent, is the modulus.

[0011] Preferably, the data management layer includes: Distributed key management module, used to generate and distribute encryption keys and ensure that key fragments of multiple parties are required during the decryption process; The encryption algorithm management module is used to manage and apply encryption algorithms.

[0012] Preferably, the distributed key management module supports a distributed reconstruction process of key fragments, and uses Lagrange interpolation method to recover the key, and the reconstruction formula is: ; in, For each party's key fragment, The identifier of the party.

[0013] Preferably, the user interface layer includes: The identity authentication module is used to complete user identity authentication without revealing the user's private key; Collaborative analysis module, which is used to conduct joint data analysis among multiple parties and supports secure collaborative analysis through multi-party computing technology without sharing the original data; The identity authentication module of the user interface layer includes an identity authentication mechanism based on zero-knowledge proof, which can complete identity authentication without leaking the user's private key. The key calculation formula of its zero-knowledge proof protocol is: ; in, is the generator, is a private random number, is the public commitment value of the prover, is a large prime number.

[0014] Preferably, the incentive and equity distribution layer includes: Contribution evaluation module, used to evaluate the contribution of each participant in data sharing and model training; The token allocation module allocates token rewards according to the predetermined rules based on the results of contribution evaluation, and encourages all participants to actively participate in data sharing and collaboration; The contribution evaluation module of the incentive and equity distribution layer evaluates the contribution of each participant based on the game theory model and distributes tokens according to the game theory. The profit calculation formula is: ; in, For the benefit of the participants, For the reward, Costs to the participants.

[0015] Preferably, the blockchain layer is connected to the remaining layers through a network, the federated learning layer relies on the blockchain layer for model training and parameter aggregation, the data management layer is connected to the blockchain layer and the federated learning layer through a network, the user interface layer is the entrance to the system and connects users to other layers through a network, and the incentive and equity distribution layer is connected to the blockchain layer and the federated learning layer through a network.

[0016] A financial data sharing method based on blockchain and federated learning, used in a financial data sharing system based on blockchain and federated learning, comprising the following steps: Step 1: The user authenticates his / her identity through the user interface layer. The system uses encryption technology to verify the user's identity and confirms the user's operation authority through the smart contract of the blockchain layer to ensure the accuracy of authority allocation; Step 2: On the user's local computer, the data management layer encrypts the data to be shared and uses the distributed key management module to generate and distribute key fragments to ensure that the key can only be reconstructed when specific conditions are met. Step 3: Each participant independently conducts model training on its own data set. The updated model parameters are encrypted and transmitted to the parameter aggregation module of the federated learning layer. Step 4: The parameter aggregation module receives and integrates the encrypted model parameters of each participant, uses the aggregation algorithm to form the parameters of the global model, and applies privacy protection measures to ensure data security; Step 5: The blockchain layer records all data access, model training, parameter transmission and aggregation operations involved, and ensures the consistency of operations between different blockchain networks through the cross-chain communication module in the multi-chain architecture; Step 6: The incentive and equity distribution layer evaluates the contribution of each participant based on their contribution, and distributes rewards according to predetermined rules through the token allocation module. All processes are recorded and executed through blockchain smart contracts.

[0017] The present invention provides a financial data sharing system and method based on blockchain and federated learning. It has the following beneficial effects: 1. The present invention adopts blockchain technology and federated learning technology to realize distributed model training and data sharing without sharing the original data. The blockchain layer records all data operation logs and permission management information, and ensures the security and compliance of data access through smart contracts.

[0018] 2. The present invention supports the integration and collaboration of multiple blockchain networks through a multi-chain architecture and a cross-chain communication module, and realizes the secure transmission of data and status between different blockchain networks. The adopted Merkle proof mechanism ensures the consistency and integrity of data in cross-chain transmission. The multi-chain architecture improves the scalability of the system, allowing the system to adapt to more application scenarios and participants, while enhancing the flexibility of the system to cope with complex financial data sharing needs.

[0019] 3. The present invention uses the contribution evaluation module and token allocation module to scientifically and rationally evaluate the contribution of each participant in data sharing and model training based on the game theory model, and allocates token rewards according to the evaluation results and rules. This mechanism ensures that each participant can get corresponding rewards for their efforts, and encourages all parties to actively participate in the collaboration and data sharing of the system.

[0020] 4. The data management layer of the present invention adopts distributed key management technology, generates and distributes encryption keys through the Shamir secret sharing algorithm, and ensures the security and reliability of key management. The distributed key management module prevents single point failure problems, and through the distributed reconstruction technology of key fragments, ensures the security of data during decryption. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a system framework diagram of the present invention; Figure 2 This is a schematic diagram of the blockchain layer framework of the present invention; Figure 3 This is a schematic diagram of the federated learning layer framework of the present invention; Figure 4 This is a schematic diagram of the data management layer framework of the present invention; Figure 5 It is a schematic diagram of the user interface layer framework of the present invention; Figure 6 This is a schematic diagram of the incentive and equity allocation layer framework of the present invention; Figure 7 It is a schematic diagram of the method steps of the present invention. DETAILED DESCRIPTION

[0022] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. Example

[0023] Please refer to the attached Figure 1 , the embodiment of the present invention provides a financial data sharing system based on blockchain and federated learning, the blockchain layer is connected to the other layers through a network, the federated learning layer relies on the blockchain layer for model training and parameter aggregation, the data management layer is connected to the blockchain layer and the federated learning layer through a network, the user interface layer is the entrance of the system and connects users to other layers through a network, and the incentive and equity distribution layer is connected to the blockchain layer and the federated learning layer through a network: Specifically, the blockchain layer is the core infrastructure of the system, responsible for managing and recording all operation logs, data access records and permission management information during data sharing, model training and parameter aggregation. The distributed ledger technology of the blockchain layer ensures the transparency and immutability of data operations, and automatically executes data sharing agreements through smart contracts to ensure that all parties exchange and collaborate on data in a compliant and trusted environment. This layer is connected to the federated learning layer, data management layer, user interface layer and incentive and equity distribution layer through the network, providing a trusted operating environment for each layer.

[0024] The federated learning layer relies on the trust mechanism and data recording function of the blockchain layer to achieve distributed model training and parameter aggregation. Each participant independently trains the local model through the federated learning layer without sharing the original data, and then transmits the encrypted model parameters to the blockchain layer through the network for recording, and passes them to the parameter aggregation module of the federated learning layer. This module uses the weighted average algorithm and differential privacy technology to aggregate the global model to ensure the accuracy of the model and the privacy of the data of the participants. The design of the federated learning layer enables all parties to share the results of collaborative analysis while maintaining data sovereignty, improving data utilization efficiency and the reliability of analysis results.

[0025] The data management layer is connected to the blockchain layer and the federated learning layer through the network, and is responsible for encrypting, access controlling, and privacy protecting all data in the system. This layer integrates a distributed key management module, uses advanced encryption algorithms to protect data, and generates and manages encryption keys through the Shamir secret sharing algorithm to ensure the secure distribution and reconstruction of keys. The existence of the data management layer ensures the security of data transmission and storage in the system, and at the same time, through collaboration with the blockchain layer, ensures the traceability and compliance of all data operations.

[0026] As the entrance to the system, the user interface layer connects users to other levels of the system through the network, providing users with interactive functions such as identity authentication, data sharing, and model training participation. After being verified by the identity authentication module, users can access data and resources in the operating system according to their permissions. The user interface layer is combined with the smart contract of the blockchain layer to ensure the legality and transparency of all user operations. At the same time, the user interface layer provides collaborative analysis functions, allowing multiple users to conduct joint analysis in a secure environment, thereby improving the depth and breadth of data analysis.

[0027] The incentive and equity distribution layer is connected to the blockchain layer and the federated learning layer through the network. It is responsible for evaluating the contribution of each participant in data sharing and model training, and distributing corresponding incentives based on the evaluation results. This layer quantifies the contribution of each party through the contribution evaluation module, and uses the game theory model to ensure the fairness and rationality of the distribution. The token distribution module uses smart contracts to automatically execute reward distribution based on the contribution, incentivizing all parties to continue to participate in the system's collaboration and data sharing, forming a virtuous circle. Through the setting of the incentive and equity distribution layer, the system promotes multi-party participation and enhances the overall vitality and sustainable development capabilities of the system.

[0028] Please refer to the attached Figure 2 ,The blockchain layer is used to record all operation logs, data access records and permission management information in the data sharing process, and execute the data sharing agreement through smart contracts; include: Multi-chain architecture module to support the integration and collaboration of multiple blockchain networks; Cross-chain communication module, which is used to achieve secure transmission of data and status between different blockchain networks and ensure data consistency through cross-chain communication protocols; The multi-chain architecture module in the blockchain layer is used to support multi-party data interaction and achieve consistency in data sharing and smart contract execution between different blockchain networks through cross-chain communication protocols. The calculation formula of the Merkle proof mechanism used in cross-chain communication is: ; in, Indicates that a hash operation is performed on the data. Indicates string concatenation; Specifically, the blockchain layer is the core component of the financial data sharing system, responsible for recording and managing all operation logs, data access records, and permission management information during the data sharing process to ensure the transparency and immutability of data operations. The blockchain layer automatically executes the data sharing agreement through smart contracts to ensure that all parties exchange and collaborate under pre-set rules, thereby ensuring the fairness and reliability of the system. In addition, the blockchain layer includes a multi-chain architecture module and a cross-chain communication module to meet the integration and collaboration requirements in a multi-blockchain environment. The multi-chain architecture module is used to support the integration and collaboration of multiple blockchain networks and solve the limitations of a single chain network in terms of performance, scalability, and data isolation. Through this module, the system can integrate different blockchain networks, allowing multiple financial institutions to efficiently and securely interact with data in a cross-chain environment while maintaining their own data privacy. The cross-chain communication module further ensures the secure transmission of data and status between different blockchain networks. Through the cross-chain communication protocol, the system can achieve data consistency and synchronous execution of smart contracts between multiple chains. To ensure the consistency and security of data during cross-chain transmission, the cross-chain communication module adopts the Merkle proof mechanism, which generates a unique Merkle root hash value by hashing the data blocks and concatenating the results, thereby ensuring the integrity and immutability of the data. The blockchain layer builds a multi-chain integrated financial data sharing platform through the collaborative work of the above modules and mechanisms, providing a highly secure, reliable and scalable technical foundation for multi-party data interaction.

[0029] Please refer to the attached Figure 3 , a federated learning layer for distributed model training and parameter sharing among multiple financial institutions; include: Model training module, which is used to perform distributed model training and model updates without sharing the original data; Parameter aggregation module, used to collect and aggregate model parameters uploaded by various participants; The model training module of the federated learning layer includes a weighted average algorithm based on multi-party participation, which is used to perform model training and parameter updates without sharing the original data. The calculation formula of the weighted average algorithm is: ; in, is the weight of the global model, For the The model weights of the participants, For the The amount of data for each participant; The parameter aggregation module combines differential privacy technology to protect the data privacy of each participant by introducing noise in the parameter update process. The noise introduction formula is: ; in, are the updated model parameters, is the learning rate, is the gradient of the loss function, is Gaussian noise.

[0030] The parameter aggregation module uses a homomorphic encryption algorithm to transmit encrypted model parameters between participants to ensure privacy and security during parameter transmission. The homomorphic encryption formula is: ; in, is the model parameter The encryption result is is the public key exponent, is the modulus.

[0031] Specifically, the federated learning layer is the core computing module of the financial data sharing system, which is specifically used for distributed model training and parameter sharing among multiple financial institutions, aiming to improve the accuracy and generalization ability of the model while strictly protecting the data privacy of each institution. The federated learning layer consists of a model training module and a parameter aggregation module. The model training module allows each participant to independently perform distributed model training and updates based on their own local data sets without sharing the original data. The process is implemented through a weighted average algorithm, which ensures the security and independence of the data of each party while ensuring the overall performance of the model. The weighted average algorithm aggregates the model weights of each participant, weights them according to the proportion of the data volume of each party, and finally generates the weight of the global model.

[0032] The parameter aggregation module is responsible for collecting and aggregating the model parameters uploaded by each participant to ensure the high quality and security of the final model. To further protect the data privacy of each participant, the parameter aggregation module combines differential privacy technology to prevent speculation on the data of a single participant by introducing noise during the parameter update process.

[0033] In addition, to ensure privacy and security during parameter transmission, the parameter aggregation module uses a homomorphic encryption algorithm. This algorithm allows parameter operations to be performed in an encrypted state, ensuring that the encryption state is always maintained during parameter transmission and aggregation to prevent data leakage. Through encryption, the system ensures that even intermediate nodes in the transmission process cannot decrypt and obtain model parameters, effectively improving the security and privacy protection level of data transmission.

[0034] Please refer to the attached Figure 4 ,Data management layer, used to manage data encryption, access control and privacy protection; include: A distributed key management module for generating and distributing encryption keys and ensuring that at least key fragments of multiple parties are required during the decryption process; Encryption algorithm management module, used to manage and apply encryption algorithms; The distributed key management module in the data management layer uses the Shamir secret sharing algorithm to generate and distribute encryption keys to ensure that at least k key fragments are required during the decryption process. The key generation formula is: ; in, is the generated key fragment polynomial, is the secret value, is a randomly generated coefficient; The distributed key management module supports the distributed reconstruction process of key fragments and uses the Lagrange interpolation method to recover the key. The reconstruction formula is: ; in, For each party's key fragment, The identifier of the party.

[0035] Specifically, the data management layer is a key security module in the financial data sharing system, responsible for encrypting, access controlling, and privacy protection of all data in the system, ensuring that the confidentiality and integrity of data are strictly maintained during data sharing and collaborative analysis. The data management layer includes a distributed key management module and an encryption algorithm management module, and the submodules work together to provide strong data security management.

[0036] The distributed key management module protects the confidentiality of data by generating and distributing encryption keys. To enhance the security and anti-attack capability of key management, the system uses the Shamir secret sharing algorithm to generate and distribute encryption keys. The algorithm divides the master key into multiple key fragments and distributes the fragments to different participants, ensuring that k key fragments are required to restore the original key during the decryption process, preventing a single participant from having full decryption authority. The mechanism enhances the system's anti-destruction capability and security, preventing a single participant from decrypting data alone.

[0037] The encryption algorithm management module is responsible for managing and applying various encryption algorithms to adapt to different types of data and security requirements. This module provides standard symmetric encryption and asymmetric encryption algorithms, and can select the most appropriate encryption method according to specific scenarios, such as homomorphic encryption, hash encryption, etc., to ensure the security of data during transmission and storage. The encryption algorithm management module is responsible for regularly updating and reviewing the security of encryption algorithms to ensure that the system always uses the most advanced and secure encryption technology.

[0038] Please refer to the attached Figure 5 , the user interface layer is used for interactive operations between financial institutions for data sharing and collaborative analysis; include: The identity authentication module is used to complete user identity authentication without revealing the user's private key; Collaborative analysis module, which is used to conduct joint data analysis among multiple parties and supports secure collaborative analysis through multi-party computing technology without sharing the original data; The identity authentication module of the user interface layer includes an identity authentication mechanism based on zero-knowledge proof, which can complete identity authentication without leaking the user's private key. The key calculation formula of its zero-knowledge proof protocol is: ; in, is the generator, is a private random number, is the public commitment value of the prover, is a large prime number.

[0039] Specifically, the identity authentication module is one of the core components of the user interface layer, responsible for verifying the user's identity without revealing the user's private key. This module uses an identity authentication mechanism based on zero-knowledge proof, allowing users to prove the legitimacy of their identity to the system without disclosing their private keys. The design of the zero-knowledge proof protocol makes it impossible for even the verifier to obtain the user's private key or other sensitive information, significantly improving the security of the identity authentication process.

[0040] The collaborative analysis module is an important component of the user interface layer, which is specifically used to support joint data analysis among multiple parties. This module uses multi-party computing technology to allow all parties to jointly participate in data analysis and model training without sharing original data. The collaborative analysis module uses encryption technology and security protocols to ensure that the data of each party remains confidential throughout the analysis process and generates aggregated analysis results. The design protects data privacy, allowing all parties to make full use of each other's data resources for in-depth analysis, thereby achieving more accurate financial forecasting and risk assessment.

[0041] The user interface layer provides a safe, reliable and efficient operating platform for system users through the collaborative work of the identity authentication module and the collaborative analysis module. The identity authentication module ensures that all users participating in the operation are legally verified and prevents unauthorized access; the collaborative analysis module provides users with a joint analysis environment under strict confidentiality conditions, allowing multiple financial institutions to safely share analysis results and benefit from the value-added effect of data collaboration without mutual trust. Through the user interface layer, financial institutions can more conveniently share data and conduct joint analysis, thereby promoting data collaboration and innovative applications in the financial industry.

[0042] Please refer to the attached Figure 6 , the incentive and equity distribution layer is used for equity distribution and incentive mechanism management based on contribution.

[0043] include: Contribution evaluation module, used to evaluate the contribution of each participant in data sharing and model training; The token allocation module allocates token rewards according to the predetermined rules based on the results of contribution evaluation, and encourages all participants to actively participate in data sharing and collaboration; The contribution evaluation module of the incentive and equity distribution layer evaluates the contribution of each participant based on the game theory model and distributes tokens according to the game theory. The profit calculation formula is: ; in, For the benefit of the participants, For the reward, Costs to the participants.

[0044] Specifically, the contribution evaluation module is the core module in the incentive and equity distribution layer, which is specifically used to evaluate the actual contribution of each participant in data sharing, model training and other activities. This module is based on the game theory model for calculation, taking into account the resources, data value, computing contribution and other factors invested by each participant in the process, and comprehensively evaluates their overall contribution to the system. Through this scientific and reasonable evaluation method, the system can quantify the contribution of each party and ensure the fairness and transparency of the distribution.

[0045] The token allocation module allocates token rewards according to the results of contribution evaluation and predetermined rules. The token allocation module ensures that each participant can obtain corresponding tokens according to their contribution. The tokens are used for exchange and purchase of services within the system and converted into other forms of rewards, further motivating participants to actively participate in the long-term development of the system. The token allocation mechanism is designed to mobilize the enthusiasm of all parties, promote more financial data sharing and collaborative analysis, and thus improve the system's operating efficiency and value creation capabilities.

[0046] Through this profit calculation method, the system can reasonably distribute the profits of all parties and ensure that the efforts of all participants can be rewarded accordingly. The introduction of the game theory model allows the system's incentive mechanism to take into account single contributions, balance the benefits of all parties in long-term cooperation, and avoid short-sighted behavior caused by single contribution evaluation.

[0047] Embodiment 2: Please refer to the attached Figure 7 , an embodiment of the present invention provides a financial data sharing method based on blockchain and federated learning, comprising the following steps: Step 1: The user authenticates his / her identity through the user interface layer. The system uses encryption technology to verify the user's identity and confirms the user's operation authority through the smart contract of the blockchain layer to ensure the accuracy of authority allocation; Step 2: On the user's local computer, the data management layer encrypts the data to be shared and uses the distributed key management module to generate and distribute key fragments to ensure that the key can only be reconstructed when specific conditions are met. Step 3: Each participant independently conducts model training on its own data set. The updated model parameters are encrypted and transmitted to the parameter aggregation module of the federated learning layer. Step 4: The parameter aggregation module receives and integrates the encrypted model parameters of each participant, uses the aggregation algorithm to form the parameters of the global model, and applies privacy protection measures to ensure data security; Step 5: The blockchain layer records all data access, model training, parameter transmission and aggregation operations involved, and ensures the consistency of operations between different blockchain networks through the cross-chain communication module in the multi-chain architecture; Step 6: The incentive and equity distribution layer evaluates the contribution of each participant based on their contribution, and distributes rewards according to predetermined rules through the token allocation module. All processes are recorded and executed through blockchain smart contracts.

[0048] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A financial data sharing system based on blockchain and federated learning, characterized in that: include: The blockchain layer uses distributed ledger technology to record operation logs, data access records, and permission management information during data sharing, and uses smart contracts to execute data sharing agreements, thereby providing a trusted collaboration foundation for the federated learning layer. The multi-chain architecture module in the blockchain layer is used to support multi-source data interaction and achieve consistency in data sharing and smart contract execution between different blockchain networks through cross-chain communication protocols. The Merkle proof mechanism calculation formula used in cross-chain communication is: ； in, Indicates that a hash operation is performed on the data. Indicates string concatenation; The federated learning layer, relying on the security of the blockchain layer, realizes distributed model training and parameter sharing among multiple financial institutions, and ensures data privacy through encryption technology and secure computing methods; The model training module of the federated learning layer includes a weighted average algorithm based on multi-party participation, which is used to perform model training and parameter updates without sharing the original data. The calculation formula of the weighted average algorithm is: ； in, is the weight of the global model, For the The model weights of the participants, For the The amount of data for each participant; The data management layer manages data encryption, access control, and privacy protection based on the needs of the blockchain layer and the federated learning layer to provide data security support; The distributed key management module in the data management layer generates and distributes encryption keys using the Shamir secret sharing algorithm to ensure that k key fragments are required during the decryption process. The key generation formula is: ； in, is the generated key fragment polynomial, is the secret value, is a randomly generated coefficient; The user interface layer serves as an interactive interface for data sharing and collaborative analysis among financial institutions. It combines blockchain and federated learning functions to achieve traceability of operations and transparency of collaboration. The incentive and equity distribution layer records the contribution evaluation results through blockchain and distributes equity and manages incentives based on them, ensuring fairness and traceability in the data sharing process.

2. A financial data sharing system based on blockchain and federated learning according to claim 1, characterized in that: The blockchain layer includes: Multi-chain architecture module to support the integration and collaboration of multiple blockchain networks; The cross-chain communication module is used to achieve secure transmission of data and status between different blockchain networks, and ensure data consistency through cross-chain communication protocols.

3. A financial data sharing system based on blockchain and federated learning according to claim 1, characterized in that: The federated learning layer includes: Model training module, which is used to perform distributed model training and model updates without sharing the original data; Parameter aggregation module, used to collect and aggregate model parameters uploaded by various participants; The parameter aggregation module combines differential privacy technology to protect the data privacy of each participant by introducing noise in the parameter update process. The noise introduction formula is: ； in, are the updated model parameters, is the learning rate, is the gradient of the loss function, is Gaussian noise.

4. A financial data sharing system based on blockchain and federated learning according to claim 3, characterized in that: The parameter aggregation module uses a homomorphic encryption algorithm to transmit encrypted model parameters between participants to ensure privacy and security during parameter transmission. The homomorphic encryption formula is: ； in, is the model parameter The encryption result is is the public key exponent, is the modulus.

5. According to claim 1, a financial data sharing system based on blockchain and federated learning is characterized in that: The data management layer includes: Distributed key management module, used to generate and distribute encryption keys and ensure that key fragments of multiple parties are required during the decryption process; The encryption algorithm management module is used to manage and apply encryption algorithms.

6. A financial data sharing system based on blockchain and federated learning according to claim 5, characterized in that: The distributed key management module supports the distributed reconstruction process of key fragments and uses the Lagrange interpolation method to recover the key. The reconstruction formula is: ； in, For each party's key fragment, The identifier of the party.

7. A financial data sharing system based on blockchain and federated learning according to claim 1, characterized in that: The user interface layer comprises: The identity authentication module is used to complete user identity authentication without revealing the user's private key; Collaborative analysis module, which is used to conduct joint data analysis among multiple parties and supports secure collaborative analysis through multi-party computing technology without sharing the original data; The identity authentication module of the user interface layer includes an identity authentication mechanism based on zero-knowledge proof, which can complete identity authentication without leaking the user's private key. The key calculation formula of its zero-knowledge proof protocol is: ； in, is the generator, is a private random number, is the public commitment value of the prover, is a large prime number.

8. A financial data sharing system based on blockchain and federated learning according to claim 1, characterized in that: The incentive and equity distribution layer includes: Contribution evaluation module, used to evaluate the contribution of each participant in data sharing and model training; The token allocation module allocates token rewards according to the predetermined rules based on the results of contribution evaluation, and encourages all participants to actively participate in data sharing and collaboration; The contribution evaluation module of the incentive and equity distribution layer evaluates the contribution of each participant based on the game theory model and distributes tokens according to the game theory. The profit calculation formula is: ； in, For the benefit of the participants, For the reward, Costs to the participants.

9. A financial data sharing system based on blockchain and federated learning according to claim 1, characterized in that: The blockchain layer is connected to the other layers through a network, the federated learning layer relies on the blockchain layer for model training and parameter aggregation, the data management layer is connected to the blockchain layer and the federated learning layer through a network, the user interface layer is the entrance to the system and connects users to other layers through a network, and the incentive and equity distribution layer is connected to the blockchain layer and the federated learning layer through a network.

10. A financial data sharing method based on blockchain and federated learning, used in a financial data sharing system based on blockchain and federated learning as claimed in any one of claims 1 to 9, characterized in that: The following steps are involved: Step 1: The user authenticates his / her identity through the user interface layer. The system uses encryption technology to verify the user's identity and confirms the user's operation authority through the smart contract of the blockchain layer to ensure the accuracy of authority allocation; Step 2: On the user's local computer, the data management layer encrypts the data to be shared and uses the distributed key management module to generate and distribute key fragments to ensure that the key can only be reconstructed when specific conditions are met. Step 3: Each participant independently conducts model training on its own data set. The updated model parameters are encrypted and transmitted to the parameter aggregation module of the federated learning layer. Step 4: The parameter aggregation module receives and integrates the encrypted model parameters of each participant, uses the aggregation algorithm to form the parameters of the global model, and applies privacy protection measures to ensure data security; Step 5: The blockchain layer records all data access, model training, parameter transmission and aggregation operations involved, and ensures the consistency of operations between different blockchain networks through the cross-chain communication module in the multi-chain architecture; Step 6: The incentive and equity distribution layer evaluates the contribution of each participant based on their contribution, and distributes rewards according to predetermined rules through the token allocation module. All processes are recorded and executed through blockchain smart contracts.

Images