Multi-level risk control system and control method based on blockchain

Abstract

The present invention discloses a multi-level risk control system and control method based on blockchain, including the following steps: constructing a risk control system including a user end and an administrator end based on blockchain technology; presetting a smart contract module on the administrator end, and constructing a repayment capacity assessment model based on the smart contract module using federated learning; the user fills in personal information through the user end to apply for a credit card, and submits the demand to the administrator end; the voucher issuer uses the distributed solution of WeIdentity to review the user's personal information; using the repayment capacity assessment model to assess the user's repayment capacity based on the personal information submitted by the user and the historical information of the administrator end; calculating the user's basic credit score based on the personal information filled in by the user, and if the user's basic credit score is higher than the credit threshold, the card is issued, otherwise the card is not issued. The present invention realizes comprehensive and accurate risk control prediction through an efficient and collaborative repayment capacity assessment prediction mechanism.

Description

Technical Field

[0001] The present invention belongs to the technical field of financial risk management and control, and specifically relates to a multi-level risk control system and a control method based on blockchain. Background Art

[0002] As the core of finance, intelligent risk control is a guarantee for licensed consumer finance companies to reduce bad debts and non-performing loans. As digital technology and the economy and society converge with unprecedented breadth and depth, through the strategic empowerment of new infrastructure, in the financial field, technology has greatly improved the coverage, access and convenience of financial services. However, it should be noted that artificial intelligence (AI) also faces limitations at the application level. When using history to predict the future, there will be strong limitations, especially due to the limitations of historical data. A single historical data is likely to cause a single sample and inaccurate predictions. Therefore, if AI is used alone to solve financial problems, there are certain risks.

[0003] In the login process of the financial system, the traditional centralized digital identity authentication has many unfavorable possibilities such as leaking user privacy, being attacked by hackers, and information forgery. At present, credit card identity authentication mainly adopts the traditional password login method, and user identity information is mainly stored in a centralized way, which is facing more and more challenges, such as easy leakage caused by database collision, privacy issues in public places, etc. Centralized identity information authentication has major hidden dangers. On the one hand, if a single point of failure occurs, a large amount of user information will be leaked. Since many users use the same account password on different websites, the risk of hackers "colliding with the database" increases. On the other hand, it also exacerbates the occurrence of identity information forgery, and theft is difficult to curb. In the centralized information storage scenario, the security of user digital identity authentication is difficult to guarantee.

[0004] In the risk monitoring link, traditional risk assessment cannot accurately assess the user's repayment risk due to the data isolation and geographical isolation of banking institutions. At present, banking institutions are isolated from data and geographically, and the assessment of user repayment ability is mainly concentrated in the banking institution where the user's account is located, but the user credit information between different institutions cannot be shared. Since the same banking institution faces a relatively single user type, the risk control detection model will not be generalized enough, which will greatly reduce the accuracy of the model assessment. In addition, different branches under the same banking institution are also geographically isolated, and user information cannot be collaboratively shared. Based on the above problems, it is difficult for banks to conduct a high-level and effective assessment of user repayment ability. Summary of the invention

[0005] In view of the above problems, the present invention proposes a multi-level risk control system and control method based on blockchain. To solve the above technical problems, the technical solutions adopted by the present invention are as follows:

[0006] A multi-level risk control system based on blockchain, comprising the following steps:

[0007] S1, build a risk control system based on blockchain technology, including user-side and administrator-side;

[0008] S2, preset the smart contract module on the administrator side, and build a repayment capacity assessment model based on the smart contract module using the federated learning method;

[0009] S3, the user fills in personal information to apply for a credit card through the user end and submits the request to the administrator end;

[0010] S4, the credential issuer reviews the user’s personal information with the help of WeIdentity’s distributed solution;

[0011] S5, using the repayment ability assessment model established in step S2 to assess the user's repayment ability based on the personal information submitted by the user and the historical information on the administrator's end;

[0012] S6, calculating the user's basic credit score based on the personal information filled in by the user. If the user's basic credit score is higher than the credit threshold, the card is issued, otherwise the card is not issued.

[0013] In step S2, the smart contract module includes the following:

[0014] Node registration module: divides the roles of learning nodes according to credit, including training nodes and leader nodes;

[0015] Local model generation module: The training node generates a local training model based on the training set and the global model. The leader node normalizes the credit calculated in the previous round and uses it as the weight of this round to aggregate the local training model to generate a local model.

[0016] Local model upload module: After the local model is generated, the leader node uploads the model parameters of the local model to the blockchain;

[0017] Global model aggregation module: aggregates the received local models to obtain a global model;

[0018] Global model parameter acquisition module: responsible for obtaining the model parameters of the global model from the blockchain;

[0019] Credit upload module: upload credit to blockchain storage;

[0020] Status query module: obtains training status information, which includes the current number of local model training rounds, model update status and credit calculation quantity.

[0021] In step S2, the use of the federated learning method to construct a repayment capacity assessment model through a smart contract module includes the following steps:

[0022] ① Collect the information of authorized credit card applicants from various banks to form a sample set, which is divided into a training set and a test set;

[0023] ②, Establish a repayment capacity preliminary review model in the genesis block of the blockchain based on the federated learning method;

[0024] ③, set the contribution threshold, initial model parameters, global aggregation round number I and initial iteration number i=1, select learning nodes according to their contribution, and use the partition clustering method to divide the selected learning nodes into local training groups;

[0025] ④, select the learning node with the highest credit in each local training group as the leader node, and the other learning nodes are training nodes. The leader nodes of all local training groups form a global aggregation group:

[0026] ⑤, according to the credit of the training node, the aggregation weight is calculated. The leader node of the local training group uses the federated learning method to aggregate the initial training model trained by the training node to obtain the local model according to the calculated aggregation weight;

[0027] ⑥, the smart contract module mobilizes the global aggregation group to aggregate the received local models to obtain the global model;

[0028] ⑦, the leader node of each local training group verifies the local model using the test set, and updates the credit of all learning nodes in the training set based on the verification accuracy and historical credit, and uploads the updated credit of the learning nodes to the blockchain;

[0029] ⑧, update the initial model parameters according to the aggregated global model, judge i<I, if it is, execute i=i+1 and return to step ④, otherwise end.

[0030] In step ③, the calculation formula of the contribution of the learning node is:

[0031]

[0032] In the formula, Contribution k represents the contribution of learning node k, D k represents the local dataset size of learning node k, N represents the total number of learning nodes in the training set, and f k Represents the local CPU frequency of learning node k.

[0033] In step ⑤, the calculation formula of the aggregation weight is:

[0034]

[0035] Where G represents the number of local training groups, represents the credit of training node k in round i-1, represents the aggregate weight of training node k in the i-th round, and N represents the total number of learning nodes in the training set.

[0036] In step ⑦, the calculation formula of the credit of the learning node is:

[0037]

[0038] In the formula, represents the credit of learning node k in the i-th round of iterative training, u represents the weight factor, represents the local model parameters of learning node k in the i-th round of iterative training, represents the global model parameters obtained by global aggregation, and represents the test set based on learning node k, δ represents the learning rate, represents the credit of learning node k in the mth round, and f(·) represents the model function.

[0039] The step S3 comprises the following steps:

[0040] S3.1, when a user needs to apply for a credit card, the user fills in the identity information through the user terminal;

[0041] S3.2. After completing the identity information, upload the corresponding certification document and submit the identity information and certification document to the administrator at the same time.

[0042] According to claim 7, the management and control method of the multi-level risk control system based on blockchain is characterized in that the identity information includes basic information, occupational information and asset status. The basic information includes the user's name, ID number, mobile phone number, household registration address, residential address, residential telephone number, gender, marital status, education level and date of birth. The occupational information includes the user's company full name, position, company address, company type, company email address, company telephone number and industry. The asset status module includes the user's salary, real estate and car property.

[0043] The calculation formula of the user's basic credit score is:

[0044]

[0045] In the formula, B b襘serepresents the user's basic credit score, r represents the repayment ability score, R represents whether the user has the repayment ability. When R=1, it means that the user has the repayment ability. When R=0, it means that the user does not have the repayment ability. A represents the user's age, a represents the age score, J represents whether the user is married. When J=1, it means that the user is married. When J=0, it means that the user is not married. j represents the marriage score. E represents the user's education level. e represents the education score. S represents the user's salary. s represents the salary score. H represents the number of properties owned. h represents the property score. C represents the number of vehicles owned. c represents the vehicle score.

[0046] A multi-level risk control system based on blockchain, comprising a user end and an administrator end, wherein the user end comprises an identity information editing module for recording user identity information and a credential review module for collecting identity information certification documents of the user, and both the identity information editing module and the credential review module are connected to the administrator end;

[0047] The administrator terminal includes the following:

[0048] Credential review module: Use WeIdentity’s distributed solution to review users’ personal information;

[0049] Smart contract module: used to monitor the process of the repayment capacity assessment module to build a repayment capacity assessment model;

[0050] Repayment ability assessment module: The repayment ability assessment model can be generated by calling the smart contract module, and the repayment ability of users who have passed the credential review can be assessed based on the repayment ability assessment model and the personal information filled in by the user. User basic credit score review module: After the repayment ability assessment is passed, the user's basic credit score is calculated based on the personal information filled in by the user, and an assessment is made as to whether the card issuance action should be executed.

[0051] Beneficial effects of the present invention:

[0052] The present invention uses blockchain as the infrastructure for storing identities, which can realize distributed storage of information such as user identities, and at the same time separate the right to use data from the ownership, thereby supporting identity management and use controlled by users' authorization; focusing on solving credit card risk control problems in a fine-grained manner with blockchain technology to ensure the security of credit card applications; establishing a repayment ability assessment model based on blockchain, conducting efficient and collaborative assessment of users' repayment abilities, and realizing comprehensive and accurate risk control predictions. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0054] Figure 1 It is a schematic diagram of the process of the control method of the present invention.

[0055] Figure 2 Schematic diagram of the training of the repayment capacity assessment model.

[0056] Figure 3 This is a schematic diagram of the interface of the identity information editing module.

[0057] Figure 4 This is a schematic diagram of the interface of the identity authentication module.

[0058] Figure 5 This is a schematic diagram of the interface of the repayment capacity assessment module.

[0059] Figure 6 This is a schematic diagram of the interface of the credit score assessment module.

[0060] Figure 7 This is the architecture diagram of the financial regulatory module.

[0061] Figure 8 It is a four-group banking alliance chain architecture based on FISCO BCOS.

[0062] Fig. 9 This is a schematic diagram of the WeBASE-Front interface.

[0063] Fig.10 This is a schematic diagram of the WeBASE-Web management interface.

[0064] Fig.11 Execute the flow chart for contract deployment.

[0065] Fig.12 This is an architectural diagram of the wind control system of the present invention.

[0066] Fig.13 Schematic diagram of the execution flow of the wind control system of the present invention. DETAILED DESCRIPTION

[0067] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments 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.

[0068] Example 1: A management and control method for a multi-level risk control system based on blockchain, such as Figure 1 and Fig.13 As shown, the following steps are included:

[0069] S1, build a risk control system based on blockchain technology, including user-side and administrator-side;

[0070] When a user needs to apply for a credit card, he or she fills in the identity information, uploads relevant supporting documents and submits them through the user terminal. The bank reviews the user through the administrator terminal to determine whether to issue the credit card.

[0071] S2, preset the smart contract module on the administrator side, and build a repayment capacity assessment model based on the smart contract module using the federated learning method;

[0072] The smart contract module includes a node registration module, a local model generation module, a local model upload module, a global model aggregation module, a global model parameter acquisition module, a credit upload module and a status query module, wherein the calling functions corresponding to each module are as follows:

[0073] Register(): Node registration, register and divide the roles of each learning node in the blockchain, including training nodes and leader nodes. Each local training group contains a leader node. The selection of leader nodes and training nodes is determined according to the credit of the learning node. The node with the highest credit in each local training group is the leader node.

[0074] LocalAggregation(): After registration, the training node updates local parameters based on the training set and the global model for local training. The local parameters refer to the weight parameters and bias parameters of each layer in the neural network. The global model is obtained from the blockchain through the Get_GlobalModel() function. The leader node is responsible for off-chain aggregation of the local training model, and the credit calculated in the previous round is normalized and used as the weight of this round for local aggregation to obtain the local model.

[0075] Put_Local_Model(): Local model upload. After several rounds of iterative training and local model generation, the leader node uploads the model parameters of the local model to the blockchain for later global aggregation;

[0076] GlobalAggregation(): Global model aggregation, where the learning nodes in the global aggregation group of the initial round are selected according to the contribution of the learning nodes. The learning nodes in the global aggregation group, i.e. the leader nodes, do not participate in local training;

[0077] Get_GlobalModel(): Gets the global model parameters, which is responsible for obtaining the model parameters of the global model from the blockchain for a new round of local training;

[0078] Put_Credit(): upload credit to the blockchain to store credit records for subsequent query and verification;

[0079] Get_status(): Status query, obtain the current local model training rounds, model update status, credit calculation number and other training status information off-chain, so as to judge the next round of training iteration and iteration termination conditions. After the global model update is completed, credit calculation is performed on each learning node. If the update information is not obtained, credit calculation cannot be performed. Among them, the model update status refers to whether the global model update is completed, and the credit calculation number refers to whether the credit of all learning nodes is updated.

[0080] The method of using federated learning to build a repayment capacity assessment model through a smart contract module is described. Figure 2 As shown, the following steps are included:

[0081] ① Collect the information of credit card applicants authorized by various banking institutions to form a sample set, and divide the sample set into a training set and a test set;

[0082] ②, Establish a repayment capacity preliminary review model in the genesis block of the blockchain based on the federated learning method;

[0083] ③, set the contribution threshold, initial model parameters, global aggregation round number I and initial iteration number i=1, select learning nodes according to their contribution, and use the partition clustering method to divide the selected learning nodes into local training groups;

[0084] The calculation formula of the contribution of the learning node is:

[0085]

[0086] In the formula, Contribution k represents the contribution of learning node k, D k represents the local dataset size of learning node k, N represents the total number of learning nodes in the training set, and f k Represents the local CPU frequency of learning node k.

[0087] First, determine the contribution threshold, and then add the learning nodes corresponding to the contribution greater than the contribution threshold to the collaborative learning. The initial credit of all learning nodes is the contribution, which can ensure that each local training group has learning nodes participating in the global aggregation, greatly increasing the security and stability of collaborative learning. The partition clustering method includes the DBSCAN algorithm (Density-Based Spatial clustering of Applications with Noise) or the K-means algorithm (K-means clustering algorithm).

[0088] ④, the node registration module selects the learning node with the highest credit in each local training group as the leader node, and the other learning nodes are training nodes. The leader nodes of all local training groups form a global aggregation group:

[0089] ⑤, the local model generation module calculates the aggregation weight according to the credit of the training node. The leader node of the local training group uses the federated learning method to aggregate the initial training model trained by the training node according to the calculated aggregation weight to obtain the local model;

[0090] The calculation formula of the aggregation weight is:

[0091]

[0092] Where G represents the number of local training groups, represents the credit of training node k in round i-1, Represents the aggregate weight of training node k in the i-th round.

[0093] The initial training model is obtained by the training node of the local training group obtaining global model parameters through a global model parameter acquisition module and performing local training in combination with a local data set.

[0094] ⑥ The local model upload module uploads the local model to the blockchain, and the global model aggregation module mobilizes the global aggregation group to aggregate the received local model to obtain the global model;

[0095] ⑦ The leader node of each local training group verifies the local model using the test set, and updates the credit of all learning nodes in the training set based on the verification accuracy and historical credit, and uploads the updated credit of the learning nodes to the blockchain through the credit upload module;

[0096] The calculation formula of the credit of the learning node is:

[0097]

[0098] In the formula, represents the credit of learning node k in the i-th round of iterative training, u represents the weight factor, represents the local model parameters of learning node k in the i-th round of iterative training, represents the global model parameters obtained by global aggregation, and represents the test set based on learning node k, δ represents the learning rate, represents the credit of learning node k in the mth round, and f(·) represents the model function. Represents the model validation accuracy.

[0099] ⑧ Update the initial model parameters according to the aggregated global model, and judge i<I through the state query module. If it is, execute i=i+1 and return to step ④, otherwise end;

[0100] Each banking institution is equivalent to a learning node. Since the performance of the nodes in the group is similar, the local model training time is not much different. Therefore, local training is considered to be synchronous. For each banking institution in the local training group, the initial model parameters are first obtained from the main blockchain through the smart contract module, and then local training is performed. Assume that H banking institutions are selected to join the local training group G, where the node set is {d1, d2, …, d h , …, d H}, the other banking institution personal information datasets of these nodes are {D1, D2, …, D h , …, D H}. The goal of banking institution m is to minimize d h The loss function expectation for a small batch. The training nodes in the local training group train the local model according to their original data set. After several rounds of local training, each group aggregates the trained local model parameters. The aggregate weight of each node is its credit value. The user credit value calculation of each bank institution is related to its repayment ability prediction.

[0101] S3, the user fills in personal information to apply for a credit card through the user end and submits the request to the administrator end;

[0102] The personal information includes identity information and certification documents, and step S3 includes the following steps:

[0103] S3.1, when a user needs to apply for a credit card, the user fills in the identity information through the user terminal;

[0104] The identity information includes basic information, occupational information and asset status. The basic information includes the user's name, ID number, mobile phone number, household registration address, residential address, residential phone number, gender, marital status, education level and date of birth. The occupational information includes the user's company name, position, company address, company type, company email address, company phone number and industry. The asset status module includes the user's salary, real estate and car property.

[0105] S3.2, after completing the identity information, upload the corresponding certification document and submit the identity information and certification document to the administrator at the same time;

[0106] The certification documents mentioned above refer to various types of identity certificates, work certificates or asset certification documents provided by public security agencies, government agencies, companies and other units.

[0107] S4, the credential issuer reviews the user’s personal information with the help of WeIdentity’s distributed solution, including the following steps:

[0108] S4.1, the credential issuer conducts a preliminary review of the user’s personal information;

[0109] Preliminarily review the correctness and completeness of the information filled in by the information user, and screen out applications that do not meet the requirements in advance.

[0110] S4.2, the distributed solution based on WeIdentity compares the user’s personal information and certification documents with the information on the blockchain.

[0111] WeIdentity is a distributed multi-center technical solution that can realize secure access authorization and data exchange between entity objects. This is an existing technology, and the following application only briefly introduces it. WeIdentity DID, a distributed identity identifier of the WeIdentity component, and WeIdentity Credential, a verifiable digital certificate module, are used to realize secure access authorization and data exchange between the credit card issuer, also known as the Issuer, the credit card applicant, also known as the User, and the credential issuer, also known as the Verifier. The specific process is as follows:

[0112] The credit card applicant requests the user agent to create a unique identity (WeID) on the blockchain;

[0113] ·The credit card applicant requests the issuance of a credit card from the credit card issuer;

[0114] The credit card issuer verifies the credit card applicant's ownership of WeID and verifies that it meets the unified certificate issuance requirements, and then uses WeID to issue a digital signature (Credential) to the blockchain;

[0115] Credit card applicants send credential application requests to different credential issuers based on their own circumstances and needs;

[0116] The credential issuer verifies the credit card applicant’s ownership of the WeID and then verifies the authenticity of the digital signature on the chain.

[0117] The above process ensures that the data is centered on the entity user. At the same time, operations such as entity identity, property rights confirmation, and authorization are completed on the blockchain, which is traceable, verifiable, and cannot be tampered with.

[0118] Verifiable digital credentials

[0119] The work objectives at the WeIdentity contract level mainly include two parts: WeIdentity DID smart contract is responsible for the establishment of the on-chain ID system, including the generation of DID (Distributed IDentity), also known as WeID, the generation of DIDDocument, and the reading and updating of DID on the chain. WeIdentity Authority smart contract is responsible for the management of alliance chain permissions, including the definition and control of on-chain DID roles, operations and permissions.

[0120] WeIdentity DID smart contract mainly implements the following functions:

[0121] The logic contract stores roles and operation permissions separately, and designs functions for adding, deleting, and modifying roles and permissions for external calls. The data is obtained by accessing the data contract, and the data is processed logically and written back to the data contract. The credit card applicant's data is mainly read by the certificate issuing institution, and the identity is judged to determine whether the user is trustworthy and whether the certificate can be issued.

[0122] Data contracts focus on the definition of data structure, storage of data content, and direct interfaces for reading and writing data.

[0123] The permission contract determines the role of the visitor and determines the permissions for different operations based on the judgment results. The permissions are divided into credit card issuers, credit card applicants, and voucher issuers, and different operation permissions are granted.

[0124] The evidence contract ensures that the signature information and additional information storage of different signatories do not interfere with each other. The credential issuer can call the add record interface to add additional information to the evidence, and can add signatures from different signatories to the same evidence by calling the createEvidence interface. The hash value of the credential issuer's signature is determined at the time of creation and cannot be changed to ensure security.

[0125] S5, using the repayment ability assessment model established in step S2 to assess the user's repayment ability based on the personal information submitted by the user and the historical information on the administrator's end;

[0126] By using the trained repayment ability assessment model to assess the repayment ability of credit card applicants who have passed identity authentication, we can get the result of whether the user has the repayment ability.

[0127] S6, calculating the user's basic credit score based on the personal information filled in by the user and the corresponding base score. If the user's basic credit score is higher than the credit threshold, the card is issued, otherwise the card is not issued;

[0128] The calculation formula of the user's basic credit score is:

[0129]

[0130] In the formula, B represents the user's basic credit score, r represents the repayment ability score, R represents whether the user has the repayment ability. When R=1, it means that the user has the repayment ability, and when R=0, it means that the user does not have the repayment ability. A represents the user's age, a represents the age score, J represents whether the user is married. When J=1, it means that the user is married, and when J=0, it means that the user is not married. j represents the marriage score, E represents the user's education level. The value of E can be 0, 1, 2, 3, 4 or 5, corresponding to below primary school, primary school, junior high school, high school, college, and above college, respectively. e represents the education score, S represents the user's salary, s represents the salary score, H represents the number of properties owned, h represents the property score, C represents the number of vehicles owned, and c represents the vehicle score.

[0131] In this embodiment, r=1000, which represents the credit allocation score for the user repayment ability assessment result obtained in step S5; a=20, which represents the age is divided into 5 stages, and each stage provides 20 credit allocation points; j=200, which represents the credit level of married people is higher than that of single people; e=200, which represents the credit allocation points of different education levels, and the higher the education level, the higher the credit score; s=10; h=500; c=100; therefore, the calculation formula of the user's basic credit score corresponds to:

[0132]

[0133] The corresponding credit threshold is set to 1500 points. When the credit score of the evaluation result meets the requirements, the administrator can issue the card to the user. If it does not meet the requirements, the card will not be issued, which can prevent users with too low credit scores from registering for credit cards and threatening the financial security of the bank. Whether the user has the ability to repay is an important basis for calculating the user's basic credit score, which is obtained according to the evaluation result of step S5.

[0134] S7, construct a credit supervision contract, and use the credit supervision contract to monitor the user's real-time credit score based on the user's transaction status and the user's basic credit score obtained in step S6;

[0135] The calculation formula of the user's real-time credit score is as follows:

[0136] B′ baae =B dyna +B base ;

[0137] In the formula, B′ dyna represents the user's dynamic credit score, B′ base Indicates the user's real-time credit score.

[0138] The credit supervision contract includes a transfer supervision module, a repayment supervision module, an overdue repayment supervision module, a multiple transfer module and a credit update module. A user will perform several of the first four actions over a period of time. After one month, the credit update module will be updated every month. The calling functions corresponding to the above modules are as follows:

[0139] Transfer(): Through this module, users can transfer credit card amounts. The transfer will increase the user's credit score, and the increase in the score is directly linked to the transfer amount.

[0140] repayNum(): Through this module, users can make repayments; user repayments are good credit card usage behavior for banks, so the frequency and amount of expenditures can affect the growth of credit points, and banks prefer users who frequently use credit cards for transactions within a reasonable time and with large transaction amounts, so each transaction will increase credit points, and the credit point increase is directly linked to the amount. In this application, the credit point will increase by 1% of the transaction amount for each transaction, and the part less than 1% will be deposited into the repayment amount ratio attribute to prepare for the next transaction.

[0141] When the user is a new user, the initial value of the user's dynamic credit score is zero, and when the user uses a credit card to transfer money such as consumption, the user's dynamic credit score B' dyna The calculation formula is as follows:

[0142]

[0143] In the formula, θ represents the transaction amount when the user uses a credit card for consumption transactions, P′ represents the repayment amount ratio attribute when the user repays the credit card, γ represents the repayment amount ratio attribute benchmark, and B dyna Indicates the user's dynamic credit score before using a credit card to transfer money.

[0144] The calculation formula of the repayment amount ratio attribute P′ when the user repays the credit card is as follows:

[0145] P′=[P+(θ′modγ)]mod γ;

[0146] Where P represents the repayment amount ratio attribute before the user repays the credit card, and θ′ represents the repayment amount when the user repays the credit card.

[0147] The repayment amount ratio attribute can be used to distinguish the difference between different repayment amounts of users. The repayment amount ratio attribute benchmark γ can be set to 100, 1000 or 500. In this embodiment, γ = 100, which can distinguish the subtle difference between repaying 9900 and repaying 9000. When the repayment amount ratio attribute reaches 100, the credit score increases by 1, which avoids the problem that the credit score cannot be changed due to multiple small transactions. Because the frequency and amount of expenditure can affect the growth of credit points, and banks prefer users who frequently use credit card transactions within a reasonable time and have large transaction amounts, each transaction will increase the credit score, and the amount of credit score increase is directly linked to the amount. In this application, the credit score of each transaction will increase by 1% of the transaction amount, and the part less than 1% will be stored in the ratio attribute to prepare for the next transaction.

[0148] checkOutOfDate(): This module is used to check whether the user's repayment behavior is overdue. The user's repayment time will be stored in the user data. When the user operates, the current time and the due repayment time will be compared. If the current time exceeds the due repayment time, the system will deduct the user's dynamic credit score. In this embodiment, the deduction method is based on weeks. If it is less than a week, it will be calculated as a week. 70 credit points will be deducted for each week of overdue. In order to avoid the problem of repeated deductions, the platform will store the number of weeks that have been deducted, and further deductions will be made for overdue behaviors exceeding this number of weeks.

[0149] When the user repays on time, the user's credit score is calculated as follows:

[0150]

[0151] In the formula, B dyna Indicates the user's dynamic credit score before the user repays on time.

[0152] When a user fails to repay on time, the user's credit score is calculated as follows:

[0153]

[0154] In the formula, t now Indicates the current date when the user repays the loan, t deductIndicates the date when the credit score was deducted when the user made the last repayment. dyna Indicates the user's dynamic credit score before the user fails to repay on time.

[0155] multiTrans(): This module is used to monitor whether the user has made frequent transactions in a short period of time. This is a high-risk behavior and the user's credit score will be greatly reduced.

[0156] When a user transfers money more than twice within time t, the calculation formula of the user's credit score is as follows:

[0157]

[0158] In the formula, t t_now Indicates the current timestamp, t t_trans Indicates the timestamp of the last transfer. If a user transfers multiple times within 50 seconds, 100 points will be deducted from his credit score from the second transfer. If a user transfers integer amounts multiple times in a short period of time, 200 points will be deducted from his credit score after each transfer. This formula can be used to classify frequent transactions of integer amounts in a short period of time as high-risk behavior, and this behavior will significantly reduce the user's credit score. dyna Indicates the user's dynamic credit score before time t. In this embodiment, if the user transfers multiple times within 50 seconds, then 100 points will be deducted from his credit score at the beginning of the second transfer. If the user transfers integer amounts multiple times within time t, 200 points will be deducted from his credit score after each transfer.

[0159] In order to better control risks and stop losses, users' transfer operations require a certain amount of credit points as a basis, that is, if the credit score is lower than this amount, the transfer cannot be made, thus avoiding the infringement of the interests of the bank by users' repeated violations. The transfer function of this platform will set the maximum transfer amount based on the user's current credit score. If the credit score is less than 1500, the transfer operation will be prohibited. If the credit score is between 1500 and 3000, the single transfer amount is 2500, and if the credit score is higher than 3000, the single transfer amount is increased to 5000.

[0160]

[0161] Where N transQuota Indicates the user's single transfer limit.

[0162] updateBaseMark(): function to update credit score. In order to better play the role of credit score in motivating people, the user's amount, repayment time and other permissions will be updated every month according to the user's real-time credit score. Each update will add the user's dynamic credit score and static credit score, that is, the sum of the user's basic credit score to form a new static credit score, and clear the dynamic credit score to zero. At the same time, the user's credit limit will be changed. The credit card limit becomes the static credit score modulo 500 multiplied by 1000.

[0163] The credit card limit of the user is limited according to the user's credit score. The credit card limit is calculated as follows:

[0164]

[0165] Where N quota Indicates the user's credit card limit.

[0166] Through the above supervision of user behavior, users can be better managed to reduce risks and stop losses in time, which solves the problems in the existing technology of poor diversified and hierarchical financial supervision, unreliable traditional financial transactions, users have a variety of uncivilized or fraudulent behaviors, and it is impossible to calculate credit scores for numerous fraudulent behaviors safely and reliably. It comprehensively considers the user's credit score to flexibly measure the user's abnormal behavior in a rough and detailed manner, which can have a strong incentive effect on user behavior.

[0167] In order to simulate privacy computing within multiple banks and provide authentic and reliable blockchain services, the following uses four edge computing devices Jetson Nano, a laptop and a wireless router, and builds a four-group banking institution alliance chain architecture across physical hardware based on the blockchain underlying platform FISCOBCOS, so that the alliance chain is expanded from the original storage / execution mechanism of one chain and one ledger to a storage / execution mechanism of one chain and multiple ledgers, and data isolation and confidentiality on the same chain are achieved based on the group dimension. In the multi-group architecture, the network is shared between groups, and network message isolation between each ledger is achieved through network access and ledger whitelist. Each business group runs its own consensus algorithm independently, and different groups can use different consensus algorithms. Each ledger module mainly includes three layers from bottom to top: core layer, interface layer and scheduling layer. These three layers work together to ensure that a single business group can run independently and robustly.

[0168] In order to more conveniently monitor the blockchain and transactions in the blockchain deployed above, achieve better operation and maintenance of blockchain nodes, uniformly manage various user accounts in the blockchain, and uniformly and easily compile and deploy various smart contracts, and lower the threshold for using blockchain, the following deploys the WeBASE management platform to enrich the functions and experience of the blockchain system.

[0169] a. Deploy blockchain management platform WeBASE

[0170] In order to facilitate the monitoring and management of banks and their corresponding credit card users within a region and realize the function of distributed smart contract deployment, a visual node front-end (WeBASE-Front) is planned to be built for each JetsonNano. Fig. 9 As shown in the figure, since the users of each bank manage their private keys and digitally sign by themselves, there is a risk of leakage or forgetfulness, which may lead to the loss of assets in their accounts on the chain. We plan to build a WeBASE-Sign component to encrypt and store the private keys of users of each party and achieve unified management. In addition, the whitelist is used to control access rights to further improve the security of WeBASE-Sign.

[0171] like Fig.10 As shown in the figure, a WeBASE management platform (WeBASE-Web) was built on a virtual machine of another laptop to centrally manage the blockchain nodes, multiple smart contracts, platform account private keys, and certificate permissions on four Jetson Nanos. Figure 8 As shown, on the other hand, the node status and transaction behavior on the entire blockchain are monitored and displayed in a visual way.

[0172] b. Deploy distributed identity and verifiable digital credentials WeIdentity

[0173] WeIdentity includes two modules: distributed identity identification (WeIdentity DID) and verifiable digital credentials (WeIdentity Credential). In order to get rid of the reliance on single-center identity registration, identification and management in the traditional model, and to achieve on-chain identity identification while minimizing or selectively disclosing the entity information of credit card users, this project uses visual deployment to build the WeIdentity deployment tool. Using the WeIdentity deployment tool, it is planned to configure WeIdentity with the previous blockchain groups, nodes, accounts and databases, and manage the credit card user's identity WeID, the repair certificate issuing agency and the repair certificate type. In addition, the alliance chain administrator is used to deploy the WeIdentity smart contract.

[0174] c. Contract deployment

[0175] The multi-level risk control system based on blockchain mainly includes two types of smart contracts. Among them, the federated learning based on blockchain group division mainly designs a global aggregated pre-compiled contract, mainly based on the C++ language. Fig.11As shown, the precompiled contract will be called by the block execution engine, and the block validator will execute the block through the block execution engine. When the execution engine executes the block, it will determine whether to use the EVM or precompiled contract engine based on the address of the called contract. When the called contract address is an EVM contract, the execution engine will create and execute the EVM to execute the transaction; when the called contract address is a registered precompiled contract address, the execution engine executes the transaction by calling the precompiled contract interface corresponding to the address

[0176] d. Security and signature settings

[0177] Since bank agents in different regions are on a consortium chain and do not want institutions outside the consortium to obtain data on the consortium chain, access control is required for data on the consortium chain. According to the characteristics of the FISCO-BCOS consortium chain, the disk encryption method is selected to control access to node storage data. The main idea is that in the consortium's intranet environment, each institution independently encrypts the data on the node's hard disk based on the AES-256 encryption algorithm, and the access rights to all encrypted data are managed by the Key Manager. The Key Manager is a service deployed in the institution's intranet that specializes in managing node hard disk data access keys and cannot be accessed from the external network. When the node in the intranet is started, it obtains the access key for the encrypted data from the Key Manager to access its own encrypted data (the database and node private key stored locally on the node).

[0178] In order to better protect the identity privacy of credit card users of various banks when using credit cards, the platform intends to use the ring signature (BBSO4 solution) and RSA algorithm integrated in FISCO-BCOS to ensure the anonymity of user accounts. Specifically, by precompiling the contract address 0x5005, turning on the CRYPTO_EXTENSION compilation option, and recompiling the source code, and finally declaring the contract interface with a solidity contract to complete the call of the ring signature precompiled contract.

[0179] Example 2: A multi-level risk control system based on blockchain, such as Fig.12 As shown, it includes a user end and an administrator end, wherein the user end includes an identity information editing module for recording user identity information and a credential review module for collecting user identity information certification documents, and both the identity information editing module and the credential review module are connected to the administrator end.

[0180] like Figure 3 As shown, the identity information editing module includes the following three modules, all of which are connected to the administrator terminal;

[0181] Basic information module: used to record the user's name, ID number, mobile phone number, household registration address, residential address, residential telephone number, gender, marital status, education level and date of birth;

[0182] Career information module: used to record the user's company name, position, company address, company type, company email address, company phone number and industry;

[0183] Asset status module: used to record the user's salary, number of properties owned and number of vehicles owned.

[0184] Banks can review and manage users' credit card application requests through the management terminal, including the following modules:

[0185] Credential review module: Figure 4 As shown, WeIdentity's distributed solution is used to review users' personal information;

[0186] Smart contract module: used to monitor the process of the repayment capacity assessment module to build a repayment capacity assessment model; the specific structure is the same as that of Example 1;

[0187] Repayment Ability Assessment Module: Figure 5 As shown, a repayment capacity assessment model can be generated by calling the smart contract module, and based on the repayment capacity assessment model and the personal information filled in by the user, it is evaluated whether the user who has passed the credential review has the repayment capacity;

[0188] User basic credit score review module: Figure 6 As shown, after the repayment ability assessment is passed, the user's basic credit score is calculated based on the personal information filled in by the user, and an assessment is made as to whether to execute the card issuance action;

[0189] Credit supervision module: Figure 7 As shown, based on the user's basic credit score, the user's credit score is supervised according to the user's transaction behavior such as consumption, transfer, repayment, etc.

[0190] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A management and control method for a multi-level risk control system based on blockchain, characterized in that: The steps include: S1, build a risk control system based on blockchain technology, including user-side and administrator-side; S2, preset the smart contract module on the administrator side, and build a repayment capacity assessment model based on the smart contract module using the federated learning method; S3, the user fills in personal information to apply for a credit card through the user end and submits the request to the administrator end; S4, the credential issuer reviews the user’s personal information with the help of WeIdentity’s distributed solution; S5, using the repayment ability assessment model established in step S2 to assess the user's repayment ability based on the personal information submitted by the user and the historical information on the administrator's end; S6, calculating the user's basic credit score based on the personal information filled in by the user, and if the user's basic credit score is higher than the credit threshold, the card is issued, otherwise the card is not issued; The smart contract module includes the following: Node registration module: divides the roles of learning nodes according to credit, including training nodes and leader nodes; Local model generation module: The training node generates a local training model based on the training set and the global model. The leader node normalizes the credit calculated in the previous round and uses it as the weight of this round to aggregate the local training model to generate a local model. Local model upload module: After the local model is generated, the leader node uploads the model parameters of the local model to the blockchain; Global model aggregation module: aggregates the received local models to obtain a global model; Global model parameter acquisition module: responsible for obtaining the model parameters of the global model from the blockchain; Credit upload module: upload credit to blockchain storage; Status query module: obtains training status information, including the current number of local model training rounds, model update status, and credit calculation quantity; In step S2, the use of the federated learning method to construct a repayment capacity assessment model through a smart contract module includes the following steps: ① Collect the information of credit card applicants authorized by various banking institutions to form a sample set, and divide the sample set into a training set and a test set; ②, Establish a repayment capacity preliminary review model in the genesis block of the blockchain based on the federated learning method; ③, set the contribution threshold, initial model parameters, global aggregation round number I and initial iteration number i=1, select learning nodes according to their contribution, and use the partition clustering method to divide the selected learning nodes into local training groups; ④, select the learning node with the highest credit in each local training group as the leader node, and the other learning nodes are training nodes. The leader nodes of all local training groups form a global aggregation group: ⑤, the aggregation weight is calculated according to the credit of the training node. The leader node of the local training group uses the federated learning method to aggregate the initial training model trained by the training node according to the calculated aggregation weight to obtain the local model; ⑥, the smart contract module mobilizes the global aggregation group to aggregate the received local models to obtain the global model; ⑦, the leader node of each local training group verifies the local model using the test set, and updates the credit of all learning nodes in the training set based on the verification accuracy and historical credit, and uploads the updated credit of the learning nodes to the blockchain; ⑧ Update the initial model parameters according to the aggregated global model. Determine if i < I. If so, execute i = i + 1 and return to step ④. Otherwise, end.

2. The management and control method of the multi-level risk control system based on blockchain according to claim 1 is characterized in that: In step ③, the calculation formula for the contribution degree of the learning node is: In the formula, Contribution k represents the contribution of learning node k, D k represents the local dataset size of learning node k, N represents the total number of learning nodes in the training set, and f k Represents the local CPU frequency of learning node k.

3. The control method of the multi-level risk control system based on blockchain according to claim 1 is characterized in that: In step ⑤, the calculation formula for the aggregation weight is: Where G represents the number of local training groups, represents the credit of training node k in round i-1, represents the aggregate weight of training node k in the i-th round, and N represents the total number of learning nodes in the training set.

4. The method for controlling a multi-level risk control system based on blockchain according to claim 1, characterized in that: In step ⑦, the calculation formula for the credit of the learning node is: In the formula, represents the credit of learning node k in the i-th round of iterative training, u represents the weight factor, represents the local model parameters of learning node k in the i-th round of iterative training, represents the global model parameters obtained by global aggregation, and represents the test set based on learning node k, δ represents the learning rate, represents the credit of learning node k in the mth round, and f(·) represents the model function.

5. The method for controlling a multi-level risk control system based on blockchain according to claim 1, characterized in that: The step S3 includes the following steps: S3.1, When a user needs to apply for credit card issuance, the user fills in identity information through the user terminal. S3.2, After the identity information is filled in, upload the corresponding supporting documents and synchronously submit the identity information and supporting documents to the administrator terminal.

6. The method for controlling a multi-level risk control system based on blockchain according to claim 5 is characterized in that: The identity information includes basic information, occupation information, and asset situation. The basic information includes the user's name, ID number, mobile phone number, household register address, residential address, residential phone number, gender, marital status, education level, and date of birth. The occupation information includes the full name of the user's company, position, company address, company type, company email, company phone number, and industry. The asset situation module includes the user's salary, real estate, and vehicle.

7. The method for controlling a multi-level risk control system based on blockchain according to claim 1, characterized in that: The calculation formula for the user's basic credit score is: In the formula, B base represents the user's basic credit score, r represents the repayment ability score, R represents whether the user has the repayment ability. When R=1, it means that the user has the repayment ability. When R=0, it means that the user does not have the repayment ability. A represents the user's age, a represents the age score, J represents whether the user is married. When J=1, it means that the user is married. When J=0, it means that the user is not married. j represents the marriage score. E represents the user's education level. e represents the education score. S represents the user's salary. s represents the salary score. H represents the number of properties owned. h represents the property score. C represents the number of vehicles owned. c represents the vehicle score.

8. A multi-level risk control system based on blockchain, including a user end and an administrator end, characterized in that: The user terminal includes an identity information editing module for recording the user's identity information and a voucher review module for the user to collect identity information supporting documents. Both the identity information editing module and the voucher review module are connected to the administrator terminal. The administrator terminal includes the following: Voucher review module: Use the distributed solution of WeIdentity to review the user's personal information. Smart contract module: Monitor the process of constructing a repayment ability evaluation model for the repayment ability evaluation module. Repayment ability evaluation module: Through the call of the smart contract module, a repayment ability evaluation model can be generated, and based on the repayment ability evaluation model and the personal information filled in by the user, evaluate whether the user who has passed the voucher review has the repayment ability. User basic credit score review module: After the repayment ability evaluation is passed, calculate the user's basic credit score according to the personal information filled in by the user, and evaluate whether to execute the card-issuing action. The smart contract module includes the following: Node registration module: Divide the roles of the learning nodes according to credit. The roles include training nodes and leader nodes. Local model generation module: The training nodes generate local training models based on the training set and the global model. The leader nodes aggregate the local training models using the credit normalized in the previous round as the weight for this round to generate local models. Local model upload module: After the local model is generated, the leader nodes upload the model parameters of the local model to the blockchain. Global model aggregation module: Aggregate the received local models to obtain a global model. Global model parameter acquisition module: Responsible for obtaining the model parameters of the global model from the blockchain. Credit upload module: Upload the credit to the blockchain for storage. Status query module: Obtain the training status information, which includes the current local model training rounds, model update situation, and credit calculation quantity. The generation method of the repayment ability evaluation model is as follows: ①, Collect the information of credit card application users authorized by each banking institution to form a sample set, and divide the sample set into a training set and a test set; ②, Based on the federated learning method, establish a preliminary repayment ability model in the genesis block of the blockchain; ③, Set the contribution threshold, initial model parameters, global aggregation round I, and initial iteration number i = 1. Select learning nodes according to the contribution of learning nodes, and use the partition clustering method to divide the selected learning nodes into local training groups; ④, Select the learning node with the highest credit in each local training group as the leader node, and the other learning nodes are training nodes. The leader nodes of all local training groups form a global aggregation group: ⑤, Calculate the aggregation weight according to the credit of the training nodes. The leader node of the local training group uses the federated learning method to aggregate the initial training models trained by the training nodes according to the calculated aggregation weight to obtain a local model; ⑥, The intelligent contract module mobilizes the global aggregation group to aggregate the received local models to obtain a global model; ⑦, The leader node of each local training group uses the test set to verify the local model, updates the credit of all learning nodes in the training set according to the verification accuracy and historical credit, and uploads the updated credit of the learning nodes to the blockchain; ⑧, Update the initial model parameters according to the aggregated global model, judge whether i < I. If so, execute i = i + 1 and return to step ④, otherwise end.

Images