CN114626934A

CN114626934A - Block chain-based multi-level wind control system and control method

Info

Publication number: CN114626934A
Application number: CN202210118104.7A
Authority: CN
Inventors: 王晓飞; 刘铸滔; 戴子明; 仇超; 王梓蔚; 边高阳
Original assignee: Tianjin University
Current assignee: Tianjin University
Priority date: 2022-02-08
Filing date: 2022-02-08
Publication date: 2022-06-14

Abstract

The invention discloses a block chain-based multi-level wind control system and a control method, which comprises the following steps: constructing a wind control system comprising a user side and an administrator side based on a block chain technology; an intelligent contract module is preset at a manager side, and a repayment capability evaluation model is constructed by using federal learning based on the intelligent contract module; a user fills in a personal information application credit card through a user side and submits a requirement to an administrator side; the certificate issuer reviews the personal information of the user by means of the distributed solution of WeIdentity; evaluating the repayment capacity of the user according to personal information submitted by the user and historical information of an administrator terminal by using a repayment capacity evaluation model; and calculating the basic credit score of the user according to the personal information filled by the user, issuing the card if the basic credit score of the user is higher than a credit threshold value, and not issuing the card if the basic credit score of the user is not higher than the credit threshold value. The comprehensive and accurate wind control prediction is realized through an efficient and synergistic repayment capability evaluation prediction mechanism.

Description

Block chain-based multi-level wind control system and control method

Technical Field

The invention belongs to the technical field of financial risk management and control, and particularly relates to a block chain-based multi-level wind control system and a management and control method.

Background

As a core of finance, the intelligent wind control is a guarantee for reducing bad accounts and bad effects of card holding consumption financial companies. With the convergence of digital technology and the unprecedented extent and depth of economic society, the coverage, access and convenience of technology to financial services in the financial field are greatly improved by enabling new capital construction strategies. It is still observed that Artificial Intelligence (AI) also faces limitations at the application level. When history is taken for prediction in the future, strong limitation is faced, especially, the limitation of history data is faced, and the single history data is likely to cause the problems of single sample and inaccurate prediction, so if the financial problem is solved by using AI only, certain risk exists.

In the login link of a financial system, the traditional centralized digital identity authentication has a plurality of adverse possibilities of revealing user privacy, being attacked by hackers, forging information and the like; at present, the identity authentication of a credit card mainly adopts a traditional password login mode, and the identity information of a user mainly adopts a centralized storage mode, so that more and more challenges are met, such as easy disclosure caused by bumping into a warehouse, privacy problems in public places, and the like. On one hand, if a single point of failure occurs, a large amount of user information can be leaked. Since most of account numbers and passwords used by many users on different websites are the same, the risk of hacker 'bumping into the library' is increased. On the other hand, the occurrence of identity information counterfeiting behavior is also aggravated, and the embezzling behavior is difficult to be inhibited. Under the scene of centralized information storage, the safety of user digital identity authentication is difficult to guarantee.

In the risk monitoring link, the traditional risk assessment cannot accurately assess the repayment risk of the user due to data isolation and geographical isolation of a bank organization. At present, data isolation and geographical isolation exist in banking institutions, repayment capacity evaluation of users is mainly focused on the banking institution where the user account is located, but credit information of the users among different institutions cannot be shared, and the same banking institution is single in user type, so that a wind control detection model is not generalized enough, accuracy of model evaluation is greatly reduced, and different branches under the same banking institution are also geographically isolated, and user information cannot be shared in a coordinated mode. Based on the above problems, it is difficult for banks to effectively evaluate the repayment ability of users at high level.

Disclosure of Invention

Aiming at the problems, the invention provides a block chain-based multi-level air control system and a control method. In order to solve the technical problems, the technical scheme adopted by the invention is as follows:

a block chain-based multi-level wind control system comprises the following steps:

s1, constructing a wind control system comprising a user side and an administrator side based on the block chain technology;

s2, presetting an intelligent contract module at a manager end, and constructing a repayment capability evaluation model by using a federal learning method based on the intelligent contract module;

s3, the user fills in the personal information application credit card through the user terminal and submits the requirement to the administrator terminal;

s4, the certificate issuer checks the personal information of the user by means of the distributed solution of WeIdentity;

s5, the repayment ability of the user is evaluated according to the personal information submitted by the user and the historical information of the administrator side by using the repayment ability evaluation model established in the step S2;

and S6, calculating the basic credit score of the user according to the personal information filled by the user, issuing the card if the basic credit score of the user is higher than a credit threshold value, and not issuing the card if the basic credit score of the user is not higher than the credit threshold value.

In step S2, the intelligent contract module includes the following:

a node registration module: dividing the roles of the learning nodes according to the credit, wherein the roles comprise training nodes and leader nodes;

a local model generation module: the training nodes generate local training models based on the training set and the global model, and the leader node normalizes the credit calculated in the previous round and takes the normalized credit as the weight of the current round to aggregate the local training models to generate local models;

a local model uploading module: after the local model is generated, the leader node uploads model parameters of the local model to the block chain;

a global model aggregation module: aggregating the received local models to obtain a global model;

a global model parameter acquisition module: the method comprises the steps of obtaining model parameters of a global model from a block chain;

a credit upload module: uploading the credits to blockchain storage;

a state query module: and acquiring training state information, wherein the training state information comprises the number of training rounds of the current local model, model updating conditions and credit calculation number.

In step S2, the method for constructing a repayment ability evaluation model through an intelligent contract module by using a federal learning method includes the following steps:

collecting information of credit card application users authorized by users of all banking institutions to form a sample set, and dividing the sample set into a training set and a testing set;

secondly, establishing a repayment capability preliminary examination model in the created blocks of the block chain based on a federal learning method;

thirdly, setting a contribution threshold, initial model parameters, a global aggregation round number I and an initial iteration number I to be 1, selecting learning nodes according to the contribution of the learning nodes, and performing local training group division on the selected learning nodes by adopting a partition clustering method;

selecting the highest learning node in each local training group as a leader node, wherein the other learning nodes are all training nodes, and the leader nodes of all the local training groups form a global aggregation group:

fifthly, calculating an aggregation weight according to the credit of the training nodes, and aggregating the initial training models trained by the training nodes by the leader node of the local training group according to the calculated aggregation weight by using a federal learning method to obtain a local model;

sixthly, the intelligent contract module invokes a global aggregation group to aggregate the received local models to obtain a global model;

seventhly, verifying the local model by using the test set by the leader node of each local training group, updating the credit of all the learning nodes in the training set according to the verification accuracy and the historical credit, and uploading the updated credit of the learning nodes to the block chain;

and (b) updating the initial model parameters according to the aggregated global model, judging that I is less than I, if so, executing I to I +1 and returning to the step (c), otherwise, ending.

In step (c), the formula for calculating the contribution of the learning node is:

wherein, Contribution_kRepresents the contribution degree of the learning node k, D_kRepresenting the local data set size of learning node k, N representing the total number of learning nodes in the training set, f_kRepresenting the local CPU frequency of learning node k.

In the fifth step, the calculation formula of the aggregation weight is as follows:

wherein G represents the number of groups of the local training set,

representing the credit of training node k in round i-1,

and N represents the total number of the learning nodes in the training set.

In step (c), the calculation formula of the credit of the learning node is:

in the formula (I), the compound is shown in the specification,

represents the credit of the learning node k in the ith round of iterative training, u represents a weighting factor,

representing the local model parameters of the learning node k in the ith round of iterative training,

representing global model parameters obtained by global aggregation,

and

represents a test set based on learning node k, δ represents a learning rate,

represents the credit of the learning node k in the mth round, and f (-) represents the model function.

The step S3 includes the following steps:

s3.1, when the user needs to apply for issuing the credit card, the user fills in identity information through the user side;

and S3.2, uploading the corresponding certification file after the identity information is filled in, and synchronously submitting the identity information and the certification file to an administrator terminal.

The method for managing and controlling the multi-level wind control system based on the block chain according to claim 7, wherein the identity information includes basic information, professional information and asset condition, the basic information includes a user's name, identification number, mobile phone number, household address, living telephone, gender, marital status, education level and date of birth, the professional information includes a user's company title, position, company address, company type, company mailbox, company telephone and industry, and the asset condition module includes a user's salary, house property and car property.

The calculation formula of the user basic credit score is as follows:

in the formula, B_{b se}The system comprises a user basic credit score, R and C, wherein the user basic credit score represents a user basic credit score, the R represents whether the user has a repayment capability, when the R is 1, the user has the repayment capability, when the R is 0, the user does not have the repayment capability, A represents the user age, a represents an age basic score, J represents whether the user is married, when the J is 1, the user is married, when the J is 0, the user is not married, J represents a marriage basic score, E represents an education level stage in which the user is located, E represents an education basic score, S represents a user' S wage amount, S represents a wage basic score, H represents a house property quantity, H represents a house property basic score, C represents a vehicle property quantity, and C represents a vehicle basic score.

A multi-level wind control system based on a block chain comprises a user side and an administrator side, wherein the user side comprises an identity information editing module for recording user identity information and a certificate auditing module for collecting identity information certification files of users, and the identity information editing module and the certificate auditing module are both connected with the administrator side;

the administrator side comprises the following steps:

a credential review module: the personal information of the user is checked by using a distributed solution of WeIdentity;

the intelligent contract module: the process for establishing the repayment capacity evaluation model by the repayment capacity evaluation module is monitored;

repayment ability evaluation module: a repayment capability evaluation model can be generated through the calling of the intelligent contract module, and whether the user who passes the certificate examination has the repayment capability or not is evaluated based on the repayment capability evaluation model and the personal information filled in by the user; the basic credit score auditing module of the user: and after the repayment capacity evaluation is passed, calculating the basic credit score of the user according to the personal information filled in by the user, and evaluating whether to execute the card issuing action.

The invention has the beneficial effects that:

the invention utilizes the block chain as an infrastructure for storing the identity, can realize the distributed storage of the information such as the user identity and the like, and simultaneously separates the use right and the ownership of the data, thereby supporting the identity management and the use of the user authorization control; focusing on solving the credit card wind control problem by the block chain technology fine granularity, and ensuring the safety of credit card application; and establishing a repayment capability evaluation model based on the block chain, and performing efficient collaborative evaluation on the repayment capability of the user to realize comprehensive and accurate wind control prediction.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.

Fig. 1 is a schematic flow chart of a management and control method according to the present invention.

Fig. 2 is a training diagram of a repayment ability evaluation model.

Fig. 3 is a schematic interface diagram of an identity information editing module.

FIG. 4 is a schematic diagram of an interface of an identity authentication module.

FIG. 5 is a schematic diagram of an interface of the payment capability evaluation module.

FIG. 6 is a schematic diagram of an interface for a credit score evaluation module.

FIG. 7 is an architecture diagram of a financial administration module.

Fig. 8 is a FISCO BCOS based four-group banking alliance chain architecture.

FIG. 9 is a schematic representation of the WeiBASE-Front interface.

FIG. 10 is a schematic diagram of a WeiBASE-Web management interface.

FIG. 11 is a contract deployment execution flow diagram.

Fig. 12 is an architecture diagram of the wind control system of the present invention.

Fig. 13 is a schematic flow chart of the execution of the wind control system according to the present invention.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be obtained by a person skilled in the art based on the embodiments of the present invention without inventive step, are within the scope of the present invention.

Example 1: a block chain-based multi-level wind control system management and control method, as shown in fig. 1 and 13, includes the following steps:

when a user needs to apply for the credit card issuance, the user fills in identity information through the user side, uploads a related certificate and submits the certificate, and a bank checks the user through the administrator side to determine whether to issue the credit card.

the intelligent contract module comprises a node registration module, a local model generation module, a local model uploading module, a global model aggregation module, a global model parameter acquisition module, a credit uploading module and a state query module, wherein calling functions corresponding to the modules are as follows:

register (): the node registration is to register and divide roles of all learning nodes in a block chain, and comprises training nodes and leader nodes, wherein each local training group comprises one leader node, the leader nodes and the training nodes are selected according to credit of the learning nodes, and the highest signal utilization in each local training group is the leader node;

LocalAggregation (): after registration is completed, the training nodes update local parameters to perform local training based on a training set and a global model, wherein the local parameters refer to weight parameters and bias parameters of each layer in a neural network, the global model is obtained from a block chain through a Get _ GlobalModel () function, the leader node is responsible for aggregating the local training models under the chain, and the credit calculated in the previous round is normalized and then is used as the weight of the round to perform local aggregation to obtain the local model;

put _ Local _ Model (): uploading a local model, wherein after several rounds of iterative training of the local model generation, a leader node uploads model parameters of the local model to a block chain so as to perform global aggregation at a later stage;

globalggregation (): global model aggregation, wherein learning nodes in a global aggregation group of an initial round are selected according to contribution degrees of the learning nodes, and the learning nodes in the global aggregation group, namely leader nodes, do not participate in local training;

get _ GlobalModel (): acquiring global model parameters, and taking charge of acquiring model parameters of a global model from a block chain so as to perform a new round of local training;

put _ Credit (): uploading the credit to a block chain for credit record storage so as to facilitate subsequent inquiry and verification;

get _ status (): and (3) state query, namely obtaining training state information such as the number of training rounds of the current local model, model updating condition, credit calculation number and the like under a link so as to judge the training iteration and iteration termination condition of the next round, performing credit calculation on each learning node after the global model is updated, and if the updating information is not obtained, not performing credit calculation. The model updating condition refers to whether the global model is completely updated or not, and the credit calculation quantity refers to whether the credits of all the learning nodes are completely updated or not.

The method for establishing the repayment capability evaluation model through the intelligent contract module by using the federal learning method is shown in fig. 2 and comprises the following steps:

firstly, collecting information of credit card application users authorized by users under various banking institutions to form a sample set, and dividing the sample set into a training set and a testing set;

secondly, establishing a repayment capacity primary review model in the creature block of the block chain based on a federal learning method;

the calculation formula of the contribution degree of the learning node is as follows:

Firstly, determining a contribution threshold, and then adding the learning node corresponding to the contribution greater than the contribution threshold into the collaborative learning. The initial credits of all the learning nodes are contribution degrees, so that the learning nodes of each local training set can participate in global aggregation, and the safety and the stability of collaborative learning are greatly improved. The partition clustering method comprises a DBSCAN algorithm (sensitivity-Based Spatial clustering of applications with Noise) or a K-means algorithm (K-means clustering algorithm).

And fourthly, the node registration module selects the highest learning node in each local training group as a leader node, all other learning nodes are training nodes, and the leader nodes of all the local training groups form a global aggregation group:

fifthly, the local model generation module calculates an aggregation weight according to the credit of the training nodes, and the leader node of the local training group aggregates the initial training models trained by the training nodes by using a federal learning method according to the calculated aggregation weight to obtain a local model;

the calculation formula of the aggregation weight is as follows:

wherein G represents the number of groups of the local training set,

representing the credit of training node k in round i-1,

and representing the aggregation weight of the training node k in the ith round.

The initial training model is obtained by the training nodes of the local training set through obtaining global model parameters by a global model parameter obtaining module and carrying out local training by combining with a local data set.

The local model uploading module uploads the local model to the block chain, and the global model aggregation module invokes a global aggregation group to aggregate the received local model to obtain a global model;

the leader node of each local training group verifies the local model by using the test set, updates the credit of all learning nodes in the training set according to the verification accuracy and the historical credit, and uploads the updated credit of the learning nodes to the block chain through the credit uploading module;

the calculation formula of the credit of the learning node is as follows:

in the formula (I), the compound is shown in the specification,

representing global model parameters obtained by global aggregation,

and

represents a test set based on learning node k, δ represents the learning rate,

Representing the model validation accuracy.

Updating initial model parameters according to the aggregated global model, judging that I is less than I through a state query module, if I is executed to be I +1, returning to the step (r), and if not, finishing;

each bank organization is equivalent to a learning node, and due to the fact that the performance of the nodes in the group is similar, the training time difference of the local model is not large. Thus, the local training is considered to be synchronous. For the banking institutions in each local training set, initial model parameters are obtained from the master block chain through the intelligent contract module, and then local training is performed. Suppose that H banking institutions are selected to join local training group G, where the set of nodes is { d }₁，d₂，…，d_h，…，d_HThe personal information data sets of other banking institutions of the nodes are { D }₁，D₂，…，D_h，…，D_H}. The goal of banking institution m is to minimize d_hExpected for a loss function for a small batch. Training local models by training nodes in local training groups according to original data sets of the local models, aggregating the trained local model parameters by each group after several rounds of local training, and aggregating each nodeThe weight value is the credit value of the user, and the calculation of the credit value of the user of each banking institution is related to the prediction of the repayment capacity of the user.

the personal information includes identity information and a certificate, and the step S3 includes the steps of:

the identity information comprises basic information, occupation information and an asset condition, the basic information comprises the name, the identity card number, the mobile phone number, the household address, the living telephone, the gender, the marital state, the education degree and the birth date of the user, the occupation information comprises the company full name, the position, the company address, the company type, the company mailbox, the company telephone and the industry of the user, and the asset condition module comprises the salary, the house property and the car property of the user.

S3.2, uploading the corresponding certification file after the identity information is filled in, and synchronously submitting the identity information and the certification file to an administrator terminal;

the certification documents refer to various kinds of identification, working certification or asset certification documents provided by public security agencies, government agencies, companies and other units.

S4, the certificate issuer reviews the personal information of the user by means of the distributed solution of WeIdentity, including the steps of:

s4.1, the certificate issuer carries out preliminary examination on the personal information of the user;

and preliminarily checking the correctness and completeness of the information filled by the information user, and screening out the applications which do not meet the requirements in advance.

S4.2, comparing personal information and certification files of the user with information on the block chain based on the distributed solution of WeIdentity.

WeIdentity is a distributed multi-center technology solution that can implement secure access authorization and data exchange between entity objects, which is a prior art and only briefly introduced in the following application. The distributed identity identification (WeIdentity DID) of a WeIdentity component and a verifiable digital certificate (WeIdentity Credential) module are utilized to realize the safe access authorization and data exchange among a credit card issuing organization, namely an Issuer, a credit card applicant, namely a User, and a certificate Issuer, namely a Verifier. The specific process is as follows:

the credit card applicant requests the user agent to create a unique identification (WeiD) on the blockchain;

the credit card applicant requests the issuance of a credit card from the credit card issuer;

the credit card issuing authority verifies the ownership of the WeID by the credit card applicant, and after verifying that the WeID complies with the unified certificate issuing requirements, issues a digital signature (Credential) to the blockchain using the WeID;

the credit card applicant sends a voucher application request to different voucher issuing organizations according to the self condition and the requirement;

the credential issuer verifies the credit card applicant's ownership of the WeID, and secondly verifies the authenticity of the digital signature on the chain.

The above process ensures that the data is centered on the entity user, and the operations of entity identity, right confirmation, authorization and the like are completed on the block chain, and can be traced, verified and not tampered.

Verifiable digital certificates

The working target of the WeIdentity contract level mainly comprises two parts: the Weidentity DID intelligent contract is responsible for establishing an ID system on a chain, and specifically comprises the steps of generating a DID (distributed identity), namely a WeID, generating a DID Document, and reading and updating the DID on the chain. The Weidentity Authority intelligent contract is responsible for alliance chain Authority management, and specifically comprises definition, operation and Authority definition and control of DID roles on a chain.

The Weidentity DID intelligent contract mainly realizes the following functions:

the logic contract stores roles and operation authorities respectively, and the roles and the authorities respectively design adding and deleting modification functions for external calling. And obtaining data by accessing the data contract, performing logic processing on the data, and writing back the data contract. The data of the credit card applicant is read by a certificate issuing organization, and the identity of the credit card applicant is judged to determine whether the user is credible or not and whether the certificate can be issued or not.

Data contracts focus on the definition of data structures, the storage of data content, and the direct interface of data reads and writes.

The rights contract judges the role of the visitor and determines the rights of different operations based on the judgment result. The authority is divided into a credit card issuing organization, a credit card applicant and a certificate issuing party, and different operation authorities are given.

The vouching contract ensures that the signature information and the additional information storage of different signing parties do not interfere with each other. The credential issuer may call the add record interface to add additional information to the deposit certificate and may add signatures from different signers to the same deposit certificate by calling the createev reference interface. The Hash value signed by the credential issuer is determined at creation time and is not alterable to ensure security.

and (4) evaluating the repayment capacity of the credit card application user who passes the identity authentication by using the trained repayment capacity evaluation model, so that whether the user has the repayment capacity can be obtained.

S6, calculating the basic credit score of the user according to the personal information filled by the user and the corresponding basic score, issuing the card if the basic credit score of the user is higher than the credit threshold value, otherwise not issuing the card;

the calculation formula of the user basic credit score is as follows:

wherein B represents the basic credit score of the user, R represents the repayment capability basic score, R represents whether the user has the repayment capability or not, when R is 1, the user is provided with repayment ability, when R is 0, the user is not provided with repayment ability, A is the age of the user, a is the age base, J is whether the user is married or not, when J is 1, it indicates that the user is married, when J is 0, it indicates that the user is not married, J indicates a marital base score, E indicates an education level stage where the user is located, the value of E may be 0, 1, 2, 3, 4, or 5, corresponding to below primary school, junior middle, high middle, university, and college, respectively, E indicates an education base score, S indicates a payroll amount of the user, S indicates a payroll base score, H indicates a property base score, C indicates a vehicle-owned number, and C indicates a vehicle base score.

In this embodiment, r is 1000, which indicates the credit allocation to the result of the evaluation of the repayment ability of the user obtained in step S5; a-20, representing a 5 year old stage of age, each stage providing a credit allocation of 20; j is 200, which means that the credit level of the married person is higher than the credit allocation of the single person; e-200, representing credit scores for different education levels, the higher the education level the higher its credit score; s is 10; h is 500; c is 100; the formula for calculating the user base credit score is therefore:

the corresponding credit threshold value is set to be 1500 minutes, when the credit score of the evaluation result meets the requirement, the administrator can execute card issuing operation on the user, if the credit score does not meet the requirement, the user does not issue the card, and the situation that the financial security of the bank is threatened by the user registration credit card with too low credit score can be avoided. Whether the user has the repayment capability is an important basis for calculating the basic credit score of the user and is obtained according to the result evaluated in the step S5.

S7, constructing a credit supervision contract, and monitoring the real-time credit score of the user according to the transaction condition of the user and the basic credit score of the user obtained in the step S6 by using the credit supervision contract;

the calculation formula of the real-time credit score of the user is as follows:

B′_baae＝B_dyna+B_base；

in formula (II) to'_dynaRepresenting user dynamicsCredit score, B'_baseRepresenting the user's real-time credit score.

The credit supervision contract comprises a transfer supervision module, a repayment overdue supervision module, a multi-transfer module and a credit updating module; the user has a plurality of behaviors of the first four actions in a period of time, and the credit updating module is updated every month after one month; the calling functions corresponding to the modules are as follows:

transfer (): through the module, the user can transfer the amount of the credit card, the transfer can lead the credit score of the user to rise to some extent, and the rising score is directly hooked with the transfer amount.

repayNum (): through the module, a user can carry out repayment action; the repayment of the user is a good credit card using behavior for the bank, so the frequency and the amount of the payment can influence the increase of the credit score, and the bank prefers the user who frequently uses the credit card for transaction within a reasonable time and has a large transaction amount, so the credit score can be increased for each transaction, and the rising number of the credit score is directly linked with the amount of the credit. In the present application, the credit score increases by one percent of the transaction amount at each transaction, and less than one percent is deposited into the repayment amount proportion attribute to prepare for the next transaction.

When the user is a new user, the initial value of the user dynamic credit score is zero, and when the user uses a credit card to transfer money such as a consumption operation, the user dynamic credit score is B'_dynaThe calculation formula of (a) is as follows:

where θ represents the transaction amount when the user performs a consumption transaction using a credit card, P' represents a payment amount proportional attribute when the user performs a payment of the credit card, γ represents a payment amount proportional attribute reference, and B represents_dynaIndicating the dynamic credit score of the user before the user uses the credit card to transfer money.

The calculation formula of the repayment amount proportion attribute P' when the user repays the credit card is as follows:

P′＝[P+(θ′modγ)]mod γ；

in the formula, P represents a payment amount ratio attribute before the user performs the payment of the credit card, and θ' represents a payment amount when the user performs the payment of the credit card.

The repayment amount proportion attribute can be used for distinguishing the difference between different repayment amounts of users, the repayment amount proportion attribute reference gamma can be set to be 100, 1000 or 500, in the embodiment, gamma is 100, and the slight difference between 9900 and 9000 can be distinguished. When the proportion attribute of the repayment amount reaches 100, the credit score is increased by l, so that the problem that the credit score cannot be changed due to small-amount multiple transactions is solved. Because the frequency and the amount of expenditure can influence the increase of credit score, and banks prefer to frequently use credit cards to trade in a reasonable time and trade users with larger amount of money, the credit score is increased in each trade, the amount of the credit score increase is directly linked with the amount of money, in the application, the credit score of each trade is increased by one percent of the amount of the trade, and the part which is less than one percent is stored in the proportion attribute to prepare for the next trade.

checkOutOfDate (): the module is used for checking whether the repayment behavior of the user is overdue or not, the repayment time of the user is stored in the user data, the current time and the repayment time are compared when the user operates, and if the current time exceeds the repayment time, the system deducts the dynamic credit of the user.

When the user pays on time, the calculation formula of the credit score of the user is as follows:

in the formula, B_dynaIndicating user on-time paymentPrevious user dynamic credit score.

When the user does not pay on time, the calculation formula of the credit score of the user is as follows:

in the formula, t_nowIndicating the current date when the user paid, t_deductIndicating the date on which the credit was deducted from the last payment made by the user, B_dynaIndicating the dynamic credit score of the user before the user pays the money on time.

mutetrans (): the module is used for monitoring whether the user conducts frequent transaction of the amount in a short time, which is a high-risk behavior, and the credit score of the user is greatly reduced.

When the user transfers more than two times in the time t, the calculation formula of the credit score of the user is as follows:

in the formula, t_{t_now}Indicating the current timestamp, t_{t_trans}A timestamp indicating the last transfer. If the user transfers the money for a plurality of times within 50s, the credit score is deducted for 100 points at the second beginning of the transfer, if the user transfers the integer amount for a plurality of times within a short time, the credit score is deducted for 200 points after each transfer, the action of frequently trading the integer amount within a short time can be listed as high-risk action through the formula, the action is generated so that the credit score of the user is greatly reduced, B_dynaRepresenting the dynamic credit score of the user before time t. In this embodiment, if the user transfers multiple times within 50s, the credit score is deducted for 100 points at the second start of the transfer, and if the user transfers an integer amount multiple times within t, the credit score is deducted for 200 points after each transfer.

For better risk control and loss prevention, the transfer operation of the user needs a certain amount of credit score as a basis, namely, the transfer cannot be carried out if the credit score is lower than the credit score, so that the benefit infringement of multiple illegal behaviors of the user on a bank is avoided. The transfer function of the platform can stipulate the transfer highest amount according to the current credit score of the user, if the credit score is less than 1500, the transfer operation is forbidden, the credit score is between 1500 and 3000, the single transfer amount is 2500, the credit score is higher than 3000, and the single transfer amount is up-regulated to 5000.

In the formula, N_transQuotaRepresenting the user's single transfer amount.

updateBaseMark (): and updating the credit score function. In order to better play the incentive role of credit score, the authority of the user such as money amount, repayment time and the like is updated every month according to the real-time credit score of the user, the sum of the dynamic credit score and the static credit score of the user, namely the sum of the basic credit score of the user, is changed into a new static credit score every time of updating, the dynamic credit score is cleared, the limit of the user is changed, and the limit of the credit card is changed into the static credit score which is multiplied by 1000 after being modulo 500.

The credit card limit of the user is limited according to the credit score of the user, and the calculation formula of the credit card limit is as follows:

in the formula, N_quotaIndicating the credit limit of the subscriber.

By monitoring the user behaviors, the risk reduction and the loss prevention of the user can be better managed, and the problems that in the prior art, the diversified graded financial monitoring is poor, the traditional financial transaction is not credible, the user has various non-civilized or fraudulent behaviors, and the safe and credible credit score calculation cannot be carried out on the numerous fraudulent behaviors are solved; the credit investigation score of the user is comprehensively considered, so that the thickness of the abnormal behavior of the user can be flexibly measured, and a strong incentive effect can be generated on the behavior of the user.

In order to simulate privacy computation inside a plurality of banks and provide real and credible block chain service, a four-group bank organization alliance chain architecture crossing physical hardware is established by four edge computing devices Jetson Nano, a notebook and a wireless router based on a block chain bottom platform FISCO BCOS, so that the alliance chain is expanded from an original one-chain one-account book storage/execution mechanism to one-chain multi-account book storage/execution mechanism, and data isolation and confidentiality on the same chain are realized based on group dimension. In the multi-group architecture, the network is shared among the groups, and the network message isolation among all accounts is realized through network access and account white lists. Each service group independently runs a respective consensus algorithm, and different groups can use different consensus algorithms. Each account book module mainly comprises a core layer, an interface layer and a scheduling layer from bottom to top, and the three layers cooperate with each other to ensure that a single business group independently and robustly operates.

In order to more conveniently monitor the block chain and the transaction in the block chain deployed above and realize better operation and maintenance of the block chain link points, various user accounts in the block chain are uniformly managed, various intelligent contracts are uniformly, simply and conveniently compiled and deployed, the use threshold of the block chain is reduced, and the function and experience of the block chain system are enriched by deploying a WeiBASE management platform.

a. Deployment blockchain management platform WeBASE

In order to conveniently monitor and manage banks and credit card users corresponding to the banks in an area and realize the function of distributed intelligent contract deployment, a visual node Front (WeiBASE-Front) is constructed for each Jetsonnano. As shown in fig. 9, since the users of the banks at all parties have the risk of disclosure or forgetting when managing their private keys and performing digital signatures by themselves, thereby causing the loss of assets of their accounts on the chain, a weibase-Sign component is to be built to encrypt and store the private keys of the users at all parties, and implement unified management. In addition, the white list is used for controlling the access right to further improve the safety of the WeBASE-Sign.

As shown in fig. 10, a weibase management platform (weibase-Web) is set up on the virtual machine of another notebook computer, so that on one hand, unified management of block chain nodes, multiple intelligent contracts, platform account private keys and certificate authorities on four Jetson Nano computers is performed, as shown in fig. 8, and on the other hand, the node states and transaction behaviors on the whole block chain are monitored and displayed in a visual manner.

b. Deploying distributed identity and verifiable digital credentials Weidentity

The Weidentity includes two modules, distributed identity (Weidentity DID) and verifiable digital certificates (Weidentity Credential). In order to get rid of dependence on single-center identity registration, identification and management in a traditional mode, minimize or selectively disclose entity information of a credit card user while realizing identity identification on a chain, a WeIdentity deployment tool is built in a visual deployment mode in the project. With the WeIdentity deployment tool, WeIdentity is intended to be collocated with previous blockchain groups, nodes, accounts, and databases, and can manage the identity WeID of the credit card user, the maintenance certificate issuer, and the maintenance certificate type. In addition, the Weidentity intelligence contract is deployed with a Federation chain administrator.

c. Contract deployment

The block chain-based multi-level wind control system mainly comprises two types of intelligent contracts. The block chain group division-based federal learning mainly designs a global-aggregation precompilation contract which is mainly based on C + + language. As shown in fig. 11, the precompiled contract is called by the block execution engine, and the block verifier executes the block through the block execution engine, and determines whether to use the EVM or the precompiled contract engine according to the address of the called contract when the block is executed by the block execution engine. When the called contract address is an EVM contract, the execution engine will create and execute an 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

d. Security and signature setup

Since bank agents in different regions are on one federation chain and it is undesirable for enterprises outside the federation to obtain data on the federation chain, access control is required for data on the federation chain. According to the characteristics of the FISCO-BCOS alliance chain, the access control of the node storage data by a method of off-the-shelf encryption is selected. The main idea is that in an intranet environment of a alliance, each organization independently encrypts data of a hard disk of a node based on an AES-256 encryption algorithm, and the access rights of all encrypted data are managed through a Key Manager. Key Manager is a service which is deployed in an internal network of a mechanism and specially manages a node hard disk data access Key, and an external network cannot access the Key. When the nodes of the intranet are started, the access keys of the encrypted data are obtained from the Key Manager to access the encrypted data (the database locally stored by the nodes and the private keys of the nodes).

In order to better protect the identity privacy of each bank credit card user when using the credit card, the platform is supposed to adopt a ring signature (BBSO4 scheme) integrated in the FISCO-BCOS and an RSA algorithm to ensure the account anonymity of the user. Specifically, CRYPTO _ EXTENSION compilation option is started through precompiled contract address 0x5005, source code compilation is repeated, and finally a contract interface is declared by a solid contract to complete calling of a ring signature precompiled contract.

Example 2: a block chain-based multi-level wind control system is shown in FIG. 12 and comprises a user side and an administrator side, wherein the user side comprises an identity information editing module for recording user identity information and a certificate auditing module for collecting identity information certification files of users, and the identity information editing module and the certificate auditing module are both connected with the administrator side.

As shown in fig. 3, the identity information editing module includes the following three modules, and all the three modules are connected with the administrator terminal;

a basic information module: the system is used for recording the name, the identification card number, the mobile phone number, the household address, the living telephone, the gender, the marital state, the education degree and the birth date of the user;

a professional information module: the system is used for recording the company full name, the position, the company address, the company type, the company mailbox, the company telephone and the affiliated industry of the user;

an asset condition module: for recording the user's salaries, house ownership numbers and vehicle ownership numbers.

The bank organization can check and manage the request of the user for applying the credit card through the management terminal, and the bank organization comprises the following modules:

a credential review module: as shown in fig. 4, the personal information of the user is reviewed using the distributed solution of WeIdentity;

the intelligent contract module: the process for establishing the repayment capacity evaluation model by the repayment capacity evaluation module is monitored; the specific structure is the same as that of the embodiment 1;

repayment ability evaluation module: as shown in fig. 5, a repayment ability evaluation model can be generated through the invocation of the intelligent contract module, and whether the user who has passed the certificate examination has the repayment ability is evaluated based on the repayment ability evaluation model and the personal information filled in by the user;

the basic credit score auditing module of the user: as shown in fig. 6, after the repayment ability evaluation is passed, the basic credit score of the user is calculated according to the personal information filled by the user, and whether to execute the card issuing action is evaluated;

a credit supervision module: as shown in fig. 7, the credit score of the user is supervised on the basis of the basic credit score of the user according to the transaction behaviors of the user such as consumption, transfer, repayment and the like.

The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.

Claims

1. A control method of a multi-level wind control system based on a block chain is characterized by comprising the following steps:

2. The method for managing and controlling a multi-level block chain-based wind control system according to claim 1, wherein in step S2, the intelligent contract module comprises the following:

a local model generation module: the training nodes generate a local training model based on the training set and the global model, and the leader node normalizes the credit calculated in the previous round and takes the normalized credit as the weight of the current round to aggregate the local training model to generate a local model;

a credit upload module: uploading the credits to blockchain storage;

3. The method for managing and controlling the block chain-based multi-level wind control system according to claim 1, wherein in step S2, the method for building a repayment capability evaluation model through an intelligent contract module by using a federal learning method comprises the following steps:

thirdly, setting a contribution degree threshold value, initial model parameters, a global aggregation round number I and initial iteration times I equal to 1, selecting learning nodes according to the contribution degrees of the learning nodes, and performing local training group division on the selected learning nodes by adopting a partition clustering method;

4. The method for managing and controlling a multi-level wind control system based on a block chain according to claim 3, wherein in the step (iii), the formula for calculating the contribution of the learning node is as follows:

5. The method according to claim 3, wherein in step (v), the calculation formula of the aggregation weight is as follows:

wherein G represents the number of groups of the local training set,

representing the credit of training node k in round i-1,

and N represents the total number of the learning nodes in the training set.

6. The method according to claim 3, wherein in step (c), the calculation formula of the credit of the learning node is:

in the formula (I), the compound is shown in the specification,

representing global model parameters obtained by global aggregation,

and

7. The method for managing and controlling a multi-hierarchy wind control system based on a block chain according to claim 1, wherein the step S3 includes the steps of:

8. The method for managing and controlling the multi-level wind control system based on the block chain according to claim 7, wherein the identity information includes basic information, professional information and asset condition, the basic information includes a user's name, identification number, mobile phone number, household address, living telephone, gender, marital status, education level and date of birth, the professional information includes a user's company title, position, company address, company type, company mailbox, company telephone and industry, and the asset condition module includes a user's salary, house property and car property.

9. The method for controlling a multi-level wind control system based on a block chain according to claim 1, wherein the calculation formula of the user basic credit score is:

in the formula, B_baseThe system comprises a user basic credit score, R and C, wherein the user basic credit score represents a user basic credit score, the R represents whether the user has a repayment capability, when the R is 1, the user has the repayment capability, when the R is 0, the user does not have the repayment capability, A represents the user age, a represents an age basic score, J represents whether the user is married, when the J is 1, the user is married, when the J is 0, the user is not married, J represents a marriage basic score, E represents an education level stage in which the user is located, E represents an education basic score, S represents a user' S wage amount, S represents a wage basic score, H represents a house property quantity, H represents a house property basic score, C represents a vehicle property quantity, and C represents a vehicle basic score.

10. A multi-level wind control system based on a block chain comprises a user side and an administrator side, and is characterized in that the user side comprises an identity information editing module for recording user identity information and a certificate auditing module for collecting identity information certification files by users, and the identity information editing module and the certificate auditing module are both connected with the administrator side;

the administrator side comprises the following steps:

repayment ability evaluation module: a repayment capability evaluation model can be generated through the calling of the intelligent contract module, and whether the user who passes the certificate examination has the repayment capability or not is evaluated based on the repayment capability evaluation model and the personal information filled in by the user; the user basic credit score auditing module: and after the repayment capacity evaluation is passed, calculating the basic credit score of the user according to the personal information filled in by the user, and evaluating whether to execute the card issuing action.