Intelligent risk control system based on block chain

Abstract

The invention discloses an intelligent risk control system based on a block chain, and relates to the technical field of finance, and the system comprises a data integration module which integrates a plurality of data sources in real time, and carries out the encryption processing of transaction data through a Hash algorithm. The graph neural network model building module is used for modeling a dependency relationship between nodes through graph convolution operation, and identifying potential fraud risks between participants; the fraudulent behavior propagation chain identification module identifies core nodes and inter-acceptance influence nodes; the node feature updating module dynamically adjusts the risk level of the node; the real-time anti-fraud detection and response module obtains transaction data in real time, evaluates the risk of each transaction node, and triggers an early warning and response mechanism when a potential fraud risk is identified; and the fraud early warning module evaluates the risk level of the transaction node in real time, and triggers a response mechanism through an intelligent contract when potential fraud is identified. Through combination of the block chain and the graph neural network, the fraud risk identification, prediction and prevention and control capabilities are significantly improved.

Description

Technical Field

[0001] The present invention relates to the field of financial technology, and in particular, to an intelligent risk control system based on blockchain. Background Art

[0002] With the digitization and networking of global financial transactions, financial fraud has been increasing in areas such as cross-border payments, supply chain finance, and trading platforms. Financial institutions, enterprises, and individuals are all facing serious credit risks and economic losses. Therefore, how to effectively identify and prevent financial fraud has become a major technical problem in the current field of financial risk control.

[0003] Traditional risk control systems usually rely on rule engines and historical data analysis to identify potential risks by setting a series of fixed risk control rules. However, this method is often difficult to cope with complex and dynamically changing fraud behaviors. Especially in cross-border payment and multi-node supply chain environments, fraud behaviors may spread through multiple links and participants to form a fraud network, and fraud patterns are constantly evolving and escalating. Simply relying on traditional risk control means cannot identify complex fraud risks in a timely and accurate manner.

[0004] However, in the process of integrating graph neural networks and blockchain technology, existing risk control systems often have problems such as low data processing efficiency, difficulty in identifying the propagation chain of fraud behaviors, and insufficient real-time warning and response mechanisms. Therefore, an intelligent risk control system based on blockchain is proposed. Summary of the Invention

[0005] The purpose of the present invention is to solve the problem that in scenarios with multiple nodes and multiple parties involved such as supply chains and cross-border payments, fraud behaviors often spread through multiple links to form a complex fraud network. Existing technologies are often difficult to effectively identify these hidden fraud chains and cannot timely discover potential risk nodes and propagation paths, and a blockchain-based intelligent risk control system is proposed.

[0006] To achieve the above purpose, the present invention adopts the following technical solutions: An intelligent risk control system based on blockchain, comprising: A data integration module for real-time integrating multiple data sources through API interfaces or big data integration tools; Graph Neural Network Model Construction Module, which is used to convert the integrated data source into graph-structured data, where: each node represents a participating party, and the transaction relationship or upstream and downstream supply chain relationship between the nodes serves as an edge; each node carries feature data related to it; the graph neural network model models the relationships between nodes through graph convolution operations to obtain a graph neural network model, thereby performing in-depth learning and analysis on the dependency relationships between nodes and their adjacent nodes in the graph; the graph convolution operation includes a multi-level convolution process, where: in the first-layer convolution, each node updates its own features by aggregating the information of its direct adjacent nodes; in the second layer and higher layers of convolution, the node captures deeper dependency relationships by aggregating the information of the adjacent nodes of its adjacent nodes, gradually expanding the information propagation range layer by layer, and helping the model identify potential indirect impacts; the multi-level convolution process helps the graph neural network gradually construct a deep dependency network and form global information aggregation between nodes, supporting remote association analysis of nodes and identifying fraud behaviors propagated through indirect relationships; Fraud Behavior Propagation Chain Identification Module, which is used to identify the core nodes related to fraud behaviors in the graph, trace the propagation path of fraud behaviors, iteratively identify the directly and indirectly affected nodes through graph convolution operations, and update their risk levels based on the influence value of each node and its position in the propagation chain, and output the complete fraud propagation path, including core nodes, indirect nodes, and their risk levels and influences; Node Feature Update Module, which is used to update the features of each node according to the fraud behaviors and propagation paths identified by the Fraud Behavior Propagation Chain Identification Module; Real-time Anti-Fraud Detection and Response Module, which is used to obtain and process transaction data in real time through streaming data processing technology, evaluate the fraud risks of each transaction node, and trigger an early warning and response mechanism when potential fraud risks are identified. The processing process of the transaction data is recorded and audited through blockchain technology; Fraud Early Warning Module, which is used to based on the results of the Node Feature Update Module, combine machine learning algorithms to evaluate the risk levels of each transaction node in real time, and take measures through smart contracts when potential fraud risks are identified. The smart contract is executed through blockchain technology.

[0007] Preferably, the Graph Neural Network Model Construction Module updates the features of each node in the following way: In the graph neural network model, the features of each node are iteratively calculated through graph convolution operations. The specific update process is as follows: Among them, is the feature representation of node v in the k+1-th round, is the weight matrix, is the set of adjacent nodes of node v, is the normalization coefficient between nodes, is the activation function; the node features are dynamically updated by combining the information of adjacent nodes to capture the dependencies between nodes.

[0008] Preferably, the graph neural network model for identifying the fraud behavior of core nodes in the direct fraud chain further includes: identifying the core nodes related to fraud behavior in the graph through the graph neural network model; determining it as the implementing node of fraud through the feature analysis of the core nodes; tracing the fraud behavior of the core nodes based on the graph neural network model and finding its connection relationship with other nodes in the graph; tracing the nodes at each level through graph convolution operations and spreading the fraud behavior of the core nodes to the direct or indirect nodes associated with it; using the graph neural network model to identify indirect nodes, that is, the nodes affected through intermediate nodes; calculating the influence value of each node iteratively through graph convolution, and finally identifying all affected indirect nodes; for each node, calculating its influence in the fraud behavior propagation chain; the nodes with greater influence are considered core fraud nodes, and the nodes with smaller influence are considered indirectly affected nodes; after calculating the influence, taking the influence value of each node as an index of node features and updating its risk level according to its position in the fraud chain; based on the calculation of the graph neural network, outputting the complete fraud propagation path, including core nodes, indirect nodes, and the risk level and influence of each node.

[0009] Preferably, the fraud behavior propagation chain identification module is used to trace the propagation path of fraud behavior in the supply chain. Through the backpropagation algorithm, it traces the propagation path starting from the core node, and uses the graph neural network model to identify potential affected nodes through feature propagation between nodes. For each node, calculate its influence with the core node, and judge whether it is part of the fraud chain according to the magnitude of the influence. Calculate the influence of each node through the following formula: where, is the node to the node between the propagation coefficients, is the node at the I-th layer of features.

[0010] Preferably, the node feature update module updates the features of each node according to the fraud behaviors and propagation paths identified by the fraud behavior propagation chain identification module, specifically including: through iterative calculations of a graph neural network model, propagating the information of adjacent nodes to the target node, and gradually updating the feature data of the target node to reflect the impact of fraud behaviors; based on the results of node feature updates, dynamically adjusting the risk levels of nodes, nodes with greater impact are considered core fraud nodes, while nodes with less impact are considered indirectly affected nodes; the node feature update process updates the features of nodes and their adjacent nodes based on graph convolution operations; The dynamic adjustment of the risk levels of nodes is achieved through multiple rounds of iterative calculations of a graph neural network to gradually update node features. After each round of iteration, the risk level of a node is adjusted according to its feature update; For each node, calculate its risk score based on its distance from fraud nodes or the influence propagated in the graph neural network model, and dynamically update the risk level; The formula for updating the risk level of a node is as follows: where, is the risk level of node v in the (k + 1)-th round, is the influence of node v, is the update function of the risk score.

[0011] Preferably, the node feature update module further expands the propagation range of node features through iterative calculations of a graph convolutional network, specifically including: during the node feature update process, gradually passing the feature information of adjacent nodes to the target node, and expanding to more distant nodes through multiple rounds of iteration to ensure that the impact of fraud behaviors can be propagated throughout the supply chain network. After each round of iteration, update the node features, and the final node features reflect the potential risks of nodes in the fraud chain.

[0012] Preferably, the real-time anti-fraud detection and response module realizes real-time monitoring and response to potential fraud behaviors in the following ways: real-time monitor transaction data, obtain transaction information through streaming data processing technology, and trigger the anti-fraud response mechanism when detecting high-risk transactions; real-time evaluate the fraud risks of each transaction node through streaming computing, and trigger measures such as automatically freezing accounts and suspending transactions when necessary.

[0013] Preferably, the real-time anti-fraud detection and response module evaluates the fraud risks of each transaction node by combining machine learning algorithms. The specific evaluation method is as follows: use machine learning algorithms to train the feature data of each node, and use the trained model to perform real-time fraud risk assessment on transaction nodes. When the risk value of a node exceeds the preset threshold, trigger corresponding anti-fraud measures.

[0014] Preferably, the fraud warning module automatically executes anti-fraud measures through a smart contract, specifically including: based on the risk level of the node, automatically executing anti-fraud operations through the smart contract. When the node risk level exceeds the threshold, the smart contract automatically executes measures such as freezing the account, suspending the transaction, and notifying the relevant parties; all anti-fraud measures are executed through the smart contract.

[0015] Preferably, the multiple data sources include transaction data, supply chain data, and external data sources. The data sources include the transaction records, payment history, credit reports, and account change data of each node; the SHA-256 hash algorithm is used to encrypt each transaction data to generate a unique hash value for each transaction; the hash value is stored through the blockchain.

[0016] The present invention has the following beneficial effects: In the present invention, by using the graph structure to capture the complex relationships between nodes and the propagation paths of fraud behaviors, the system can identify cross-node fraud behaviors that cannot be detected by single-node fraud detection, providing a more comprehensive risk assessment and prevention. At the same time, through real-time node feature update and risk level adjustment, the system can dynamically evaluate and respond to risks, ensuring the timely identification and suppression of fraud behaviors. In addition, by using smart contract technology, anti-fraud measures such as freezing accounts and suspending transactions are automatically executed, greatly shortening the time of manual intervention, and improving the response speed and efficiency. The application of blockchain technology ensures the immutability and transparency of data, reduces the risk of data being tampered with or omitted, and enhances the auditability and traceability of the system. Using stream data processing technology, real-time monitoring and analysis of transaction data can quickly respond to and process fraud behaviors, providing real-time warnings.

[0017] In the present invention, not only the problem of single-node fraud detection is solved, but also the relationships between nodes are deeply analyzed to ensure the comprehensiveness and in-depth identification of risks. By automatically executing anti-fraud measures and real-time monitoring, the time of manual intervention is reduced, the work efficiency is improved, and the possibility of human errors is reduced. At the same time, through stream data processing technology and feedback loops, the system can continuously learn and adjust model parameters, improve the accuracy of fraud behavior identification, and continuously optimize the risk control strategy. The entire system provides a highly automated, real-time response, efficient, and reliable solution for fraud behavior identification and prevention in the financial field, especially in supply chain management. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a system architecture diagram of an intelligent risk control system based on blockchain proposed by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0020] As Figure 1 shown, an intelligent risk control system based on blockchain proposed by the present invention includes: A data integration module for integrating multiple data sources in real time through API interfaces or big data integration tools; A graph neural network model construction module for converting the integrated data sources into graph-structured data, where: each node represents a participant, and the transaction relationship or upstream and downstream supply chain relationship between nodes is used as an edge; each node carries feature data related to it; the graph neural network model models the relationships between nodes through graph convolution operations to obtain a graph neural network model, thereby performing in-depth learning and analysis on the dependence relationships between nodes and their adjacent nodes in the graph; the graph convolution operation includes a multi-level convolution process, where: in the first-level convolution, each node updates its own features by aggregating the information of its directly adjacent nodes; in the second-level and higher-level convolutions, the nodes aggregate the information of the adjacent nodes of their adjacent nodes, gradually expanding the scope of information propagation layer by layer to capture deeper dependence relationships, helping the model identify potential indirect impacts; the multi-level convolution process helps the graph neural network gradually construct a deep dependence network and form global information aggregation between nodes, supporting remote association analysis of nodes and identifying fraud behaviors propagated through indirect relationships; A fraud behavior propagation chain identification module for identifying the core nodes related to fraud behaviors in the graph, tracing the propagation path of fraud behaviors, iteratively identifying directly and indirectly affected nodes through graph convolution operations, and updating their risk levels based on the influence value of each node and its position in the propagation chain, and outputting a complete fraud propagation path, including core nodes, indirect nodes, their risk levels, and influence; A node feature update module for updating the features of each node according to the fraud behaviors and propagation paths identified by the fraud behavior propagation chain identification module; A real-time anti-fraud detection and response module for obtaining and processing transaction data in real time through streaming data processing technology, evaluating the fraud risks of each transaction node, and triggering an early warning and response mechanism when potential fraud risks are identified. The processing process of transaction data is recorded and audited through blockchain technology; The fraud warning module is used to, based on the results of the node feature update module, combine machine learning algorithms to conduct real-time assessment of the risk levels of each transaction node, and when potential fraud risks are identified, take measures through smart contracts, and the smart contracts are executed through blockchain technology.

[0021] The solution focuses on solving the problems of fraud identification and prevention in the financial field, especially in supply chain management. Most traditional anti-fraud technologies focus on fraud detection on a single node or transaction. However, the relevance between supply chains or multiple transaction nodes is often overlooked. When fraud occurs at one node, it often affects other related nodes directly or indirectly.

[0022] 1. Data integration module: Data source integration: Transaction data: Obtain in real time detailed information such as transaction amount, time, identities of both parties to the transaction, transaction device, and geographical location from transaction platforms (such as banks, payment gateways, e-commerce platforms, etc.). Specifically include: Obtain transaction amount, transaction time, transaction type (such as transfer, payment, refund, etc.), transaction status (such as success, failure, pending, etc.), account information of transaction participants, IP address or geographical location information of the transaction device, etc. from bank transaction data.

[0023] Obtain transaction payment information from the payment gateway, such as payment method (such as credit card, e-wallet, etc.), payment status, payment amount, payment time, payment channel information, etc.

[0024] Obtain order information from the e-commerce platform, such as order amount, order time, order status (such as paid, unpaid, refunded, etc.), information of both the buyer and the seller, product information, etc.

[0025] Supply chain data: Obtain data such as suppliers, customers, and logistics information from the supply chain management system to ensure complete records of each node (such as suppliers, customers, etc.). Specifically include: Obtain information of suppliers from the supply chain system, such as supplier name, address, contact information, credit rating, delivery history, payment history, etc.

[0026] Obtain customer information, such as customer name, address, contact information, credit rating, purchase history, payment history, etc.

[0027] Obtain logistics information, such as transportation method, transportation time, transportation status, transportation company, quantity of goods, value of goods, etc.

[0028] External data sources: Financial service data: Obtain relevant data such as credit ratings and payment histories from financial service institutions. Specifically include: Obtain data such as credit ratings, credit reports, credit histories, and credit risks of enterprises or individuals from credit rating agencies.

[0029] Obtain payment history records from payment platforms or banks, such as payment amounts, payment frequencies, payment channels, payment statuses, etc.

[0030] Data for government cloud business administration information services: Obtain enterprise registration information, tax information, etc. from government departments. Specifically include: Obtain enterprise registration information from the government's industrial and commercial registration system, such as enterprise name, registered address, registered capital, legal representative, shareholder information, business scope, etc.

[0031] Obtain the tax payment records, tax risk information, tax credit ratings, etc. of enterprises from the tax department.

[0032] Data for legal information services: Obtain legal risk information of enterprises or individuals, such as litigation, bankruptcy, etc. from legal databases. Specifically include: Obtain legal risk information such as whether enterprises or individuals are involved in litigation, legal disputes, bankruptcy, liquidation, enforcement, etc. from legal databases.

[0033] Data preprocessing: Denoising and supplementation: Use the Kalman filtering algorithm to denoise the original data to eliminate the impact of noise on data accuracy: For missing data, use linear interpolation or the average value of historical data to fill in, to ensure the integrity of the data.

[0034] Standardization: Standardize the data from different data sources to ensure the consistency of dimensions: For numerical data (such as transaction amounts, frequencies, etc.), use Z-score standardization to make the data distributed on the same scale: For categorical data (such as transaction types, participant types, etc.), use one-hot encoding or other appropriate encoding methods to make the data suitable for model training.

[0035] Ensure the integrity and immutability of the data: Use the SHA-256 algorithm to generate a unique hash value for each transaction data: Convert the transaction data into a string and ensure the consistency of the data order (such as sorting by field names).

[0036] Hash the sorted data string to generate a unique and irreversible hash value.

[0037] Blockchain storage: Record the generated hash value and related transaction data on the blockchain to ensure the security and transparency of the data: Block generation: Pack the transaction data into a new block. Each block contains multiple transaction data and their hash values.

[0038] Hash link: Connect the hash value of the new block with the hash value of the previous block to form a chain structure.

[0039] Verification and consensus: Verify and confirm the legality of the block through a consensus mechanism (such as Proof of Work - PoW, Proof of Stake - PoS, etc.).

[0040] Blockchain update: Add the verified block to the blockchain to ensure the immutability and transparency of the transaction data.

[0041] 2. Graph neural network model construction module: Graph structure construction: Convert the pre - processed data into a graph structure and construct the association relationships between nodes: Nodes: Each node represents a transaction participant, such as a supplier, a customer, a financial institution, etc. Node features include transaction records (such as transaction amount, time, frequency, etc.), credit scores, account changes, etc. Specific node features include: Transaction amount: Represents the size of the transaction amount.

[0042] Transaction time: Represents the time point when the transaction occurs.

[0043] Transaction frequency: Represents the frequency of transactions.

[0044] Credit score: The credit rating of the participant obtained from an external data source.

[0045] Account change: Represents the change in the account balance of the participant.

[0046] Rule - based relationship construction: Association rule setting: Determine the relationships between nodes through preset association rules. These rules may include: Transaction frequency: If the transaction frequency between two nodes is high within a specific time period, it is considered that there is a strong connection between them.

[0047] Transaction amount: The relationship between nodes with large and frequent transaction amounts may have a higher weight.

[0048] Participant relationship: If a node has common transaction partners or upstream - downstream relationships in the supply chain with other nodes, connections can be established through this "indirect" relationship.

[0049] Rule example: If there are more than 10 transactions with an amount greater than 5000 between Node A and Node B in the past 30 days, the system will consider that there is a strong trading relationship between A and B.

[0050] If the trading objects of Node A and Node B are similar (for example, jointly participating in the same transactions or fund transfers), they will be regarded as associated nodes.

[0051] Edge weight assignment For each pair of constructed edges, it is necessary to assign a weight to represent the association strength. This can usually be assigned based on the following methods: Transaction amount weighting: The weight of the edge can be weighted according to the transaction amount between the two nodes. The larger the amount, the higher the association weight.

[0052] Transaction frequency weighting: Edges with a high transaction frequency can be assigned a higher weight, indicating that the interaction between these two nodes is more frequent.

[0053] Transaction type weighting: If it is a high-risk transaction type (such as fund flow), a higher weight can be assigned to the edges between related nodes.

[0054] Relationship strength weighting: Set the weight according to the association relationship strength between nodes. For example, long-term partners have a higher weight.

[0055] Initialization steps of the graph model: Initialization of nodes and edges: When initially establishing the graph model, the characteristics of each node will include: Legitimacy of node behavior: For example, whether the node has any violation records (credit assessment, historical fraud records, etc.).

[0056] Degree of transaction frequency of the node: The frequency of the node's trading behavior and the interaction strength with other nodes.

[0057] Credit score of the node: Such as the credit score of the node, used to judge whether the node behavior is normal.

[0058] The attributes of each edge will include: Transaction type: Used to mark the transaction category between nodes.

[0059] Association strength: Based on the weight assignment mentioned above, it reflects the closeness of the relationship between nodes.

[0060] Initial graph structure construction: The initial graph model should be completed by constructing the basic attributes of nodes and edges. The graph model at this time can be regarded as a sparse graph because there is no direct association between every pair of nodes. Graph structure representation: Construct a graph , where: V is the set of nodes, and each node Represents a participating party (such as a user, merchant, etc.), and the node carries its characteristic data (such as transaction history, behavioral data, etc.).

[0061] Is a set of edges, representing the relationships between nodes, usually transaction relationships or upstream and downstream relationships in the supply chain.

[0062] Steps for modeling multi-level relationships between nodes: Introduction of multi-level relationships: Modeling of multi-level relationships: Excluding direct transaction relationships, further analyze and model the indirect associations between nodes. For example: Historical transaction chain: By analyzing historical transaction records, construct a historical transaction chain. For example, the relationship between node A and B is not only the current transaction, but can also be traced back to other associated nodes in their history (such as C and D).

[0063] Supply chain relationship: Node A may be a supplier of a certain product, B is its customer, and C may be a supplier of B. By tracing these upstream and downstream relationships in the supply chain, a richer graph model can be formed.

[0064] Enhancing the recognition ability of the graph model: Indirect relationships: By introducing indirect relationships between nodes, the graph model can capture nodes that do not have direct transaction relationships on the surface but have potential risk associations. For example: Cross-node indirect relationships: If node A is directly associated with node B, and at the same time node B also has a strong transaction relationship with node C, then node A and node C may be indirectly affected.

[0065] Deep optimization of the graph model: By modeling multi-level and multi-dimensional node relationships, graph neural networks can capture more complex behavioral patterns from the perspective of deep learning and identify potential fraud behaviors or abnormal patterns.

[0066] Using graph neural networks (GNNs) for node feature learning and updating: Graph convolution operation: The first layer of convolution updates the node features through the following formula: Where, Is the feature representation of node v in the k + 1 round, Is the weight matrix, Is the set of adjacent nodes of node v, Is the normalization coefficient between nodes, Is the activation function; (such as ReLU).

[0067] Second layer and higher layer convolutions (recursive information aggregation): For higher-level convolutions, nodes obtain information from farther distances through the adjacent nodes of their adjacent nodes. For example, the The convolution of layers aggregates features from the layer: This multi-layer convolution mechanism helps capture deeper dependencies, thus supporting the identification of fraud behavior propagation paths.

[0068] Iterative calculation: Through multiple rounds of iteration, node features gradually reflect the relationships between nodes and potential fraud behaviors. In each iteration, a node obtains information from its adjacent nodes, updates its own feature representation, forming a dynamically updated graph model that can reflect newly added nodes or association relationships in real time. Model output: The final output of the graph neural network is the features of each node , which contain deep learning representations of the relationships between nodes. This representation will be used in the subsequent fraud behavior propagation chain identification module.

[0069] Using historical data for supervised learning to train the model to identify fraud behaviors: Training dataset: Establish a sample dataset containing known fraudulent and non-fraudulent transactions.

[0070] Each sample in the dataset includes the feature data of the transaction (such as transaction amount, time, identity of the participating parties, etc.) and the label (whether it is fraud). Specifically, it includes: Extract features from historical transaction data, such as transaction amount, transaction time, transaction frequency, identity of transaction participating parties, transaction device information, geographical location, etc.

[0071] Based on the history of known fraudulent transactions, mark fraudulent and non-fraudulent transaction samples.

[0072] Optimization objective: Through the backpropagation algorithm, adjust the model parameters to minimize the deviation between the model output and the true labels, and maximize the accuracy of the model in identifying fraud behaviors.

[0073] Model parameter adjustment: Update the weight matrix through the gradient descent algorithm.

[0074] Determine risk points and related factors according to the business category, and make judgments according to the risk control ratio to form a rating result: Model construction: Determine risk points and related factors according to the business category, and make judgments according to the risk control ratio.

[0075] By integrating relevant data interfaces such as financial service categories, government cloud business administration information service categories, and legal information service categories, key information data that may potentially affect personal or enterprise financial risk assessment is stored in the database for rating. Specifically, it includes: Financial service data interface: Obtain credit ratings, payment history, etc. from financial service institutions.

[0076] Government Cloud Business Administration Information Service Data Interface: Obtain enterprise registration information, tax information, etc. from government departments.

[0077] Legal Information Service Data Interface: Obtain legal risk information of enterprises or individuals, such as litigation, bankruptcy, etc. from legal databases.

[0078] Rating Results: Display the rating results of the initial business review, including the ratings of enterprises / businesses.

[0079] Statistical Portrait: Automatically store and statistically analyze customer information, form an analysis and statistics page, and visualize the user portrait.

[0080] Rating Methods: Include linear scoring and grade mapping - Matrix Scoring Method: Linear Scoring: Calculate the risk score by weighted averaging according to the weights of each risk point.

[0081] Matrix Scoring: Map each risk point according to importance and influence in a matrix to determine the rating grade.

[0082] 3. Fraud Behavior Propagation Chain Identification Module: In financial risk control, the propagation chain of fraud behavior refers to a malicious account or node that affects and controls other accounts through multiple transaction behaviors, forming a behavior pattern or risk propagation chain. This chain may involve the interaction of multiple accounts and spread fraud behavior through different paths.

[0083] For example, assume that account A (the account initiating the fraud behavior) transfers funds through a series of transaction behaviors with accounts B, C, D, etc., and these accounts B, C, D, etc. may also have transaction behaviors with other accounts, resulting in the transfer of fraud behavior among accounts. At this time, accounts A, B, C, D, etc. constitute a fraud behavior propagation chain.

[0084] Identify the core nodes related to fraud behavior through the node features output by the graph neural network model, and trace the propagation path of fraud behavior. This module traverses the graph structure through depth-first search (DFS) or breadth-first search (BFS), calculates the propagation path, and identifies the influence.

[0085] Core Node Identification: Node Fraud Probability: The fraud probability of each node output by the graph neural network model , which can be used to determine whether a node is a core fraud node. Set a threshold T, and nodes exceeding this threshold are considered core fraud nodes.

[0086] The specific calculation process is as follows: Obtain the node fraud probability output by the model .

[0087] Set a threshold T. If > T, then the node is regarded as a core fraud node.

[0088] Traverse all nodes, compare their fraud probabilities with the threshold T, and identify all core fraud nodes , and add them to the list of core fraud nodes.

[0089] Trace the propagation path of fraud behavior starting from the core node. Through graph convolution operations and graph traversal algorithms, trace the propagation path of fraud behavior in the graph and identify the affected nodes.

[0090] Propagation path calculation: For each core node , use graph convolution to calculate the influence of its adjacent nodes. The calculation of influence is based on the propagation coefficient between nodes, that is, the influence between node and its adjacent node .

[0091] The influence calculation formula is as follows: where is the propagation coefficient between node and node , is the feature of node at the I-th layer.

[0092] Propagation path identification: Through the depth-first search (DFS) or breadth-first search (BFS) algorithm, starting from the core node , traverse the graph structure and identify all directly or indirectly affected nodes related to it. The specific steps are as follows: Starting from the core node , traverse the graph structure according to the DFS or BFS algorithm.

[0093] For each traversed node , calculate the influence between it and the core node .

[0094] If the influence of node exceeds a certain threshold, then this node is considered an affected node.

[0095] Add node to the fraud propagation chain and continue traversing.

[0096] 2.3 Indirect node identification and influence update Iteratively update the influence of each node through a graph neural network model, and finally identify all affected indirect nodes.

[0097] Influence update: The influence of each node is updated through graph convolution operations. The update process is as follows: Among them, is the feature of node at the (I + 1)-th layer, representing the updated influence of node

[0098] Influence magnitude judgment: Nodes with greater influence are considered core fraud nodes, while those with smaller influence are considered indirectly affected nodes.

[0099] Output: Based on the calculation of the graph neural network, the complete fraud propagation path is output, including: A list of core fraud nodes.

[0100] A list of indirectly affected nodes.

[0101] The risk level and influence of each node in the fraud behavior propagation chain.

[0102] 4. Node feature update module: According to the real-time data changes in the blockchain network, the attribute information of each node is automatically updated: Feature update: Through graph convolution operations, the information of adjacent nodes is propagated to the target node to update its feature representation. The specific steps include: Obtaining the feature data of adjacent nodes. Updating the feature representation of the target node through graph convolution operations, combining the information of adjacent nodes. Dynamically adjusting the risk level of the node according to the node feature and the influence of the fraud behavior: Risk level adjustment: Calculate the risk score of the node. The formula is as follows: Among them, is the risk level of node v at the (k + 1)-th round, is the influence of node v, is the update function of the risk score. It can be a linear or non-linear function.

[0103] Dynamically adjust the risk level of the node according to the risk score. A higher risk score indicates the key position of the node in the fraud chain. The specific adjustment methods include: Increase the risk level threshold: When the risk score of the node exceeds the preset threshold, increase its risk level.

[0104] Weight adjustment: Adjust the weight of the node in the fraud behavior propagation path according to its fraud risk score.

[0105] 5. Real-time anti-fraud detection and response module: ​Use streaming data processing technologies (such as Apache Kafka, Apache Flink) for real-time data monitoring and processing to ensure quick response: Real-time data processing: Obtain transaction data from the trading platform in real time and quickly process it through a streaming processing engine such as Kafka or Flink.

[0106] Preprocess the transaction data in real time, such as denoising, standardization, etc., to make it suitable for model input.

[0107] Perform real-time risk assessment on each trading node through a trained machine learning model: Extract features from the real-time transaction data, such as transaction amount, transaction time, identities of both parties to the transaction, device information, geographical location, etc.

[0108] Use the trained model to predict the risk score of the transaction. The specific steps include: Obtain real-time transaction data.

[0109] Extract transaction features.

[0110] Input the feature data into the trained model to predict the risk score.

[0111] When a high-risk transaction is identified, automatically trigger anti-fraud measures: Response mechanism trigger: Set a risk score threshold (such as 90). When the risk score of a certain trading node is higher than the threshold, perform the following measures: Freeze the account: Automatically freeze the high-risk account associated with the transaction.

[0112] Suspend the transaction: Suspend the trading activities associated with the node.

[0113] Notify the relevant parties: Through smart contracts or other notification mechanisms, inform the relevant regulatory agencies or responsible parties.

[0114] 6. Fraud warning module: Combine machine learning algorithms to perform real-time assessment of the risk level of each trading node: Extract features from the transaction data. Use the trained model to evaluate the feature data of each node and dynamically adjust the risk level of the node. The specific steps include: Obtain transaction data.

[0115] Extract transaction features.

[0116] Input the feature data into the trained model to predict the risk score and dynamically adjust the risk level.

[0117] Automatically execute anti-fraud measures through smart contracts: Write smart contracts on the blockchain to pre-define various anti-fraud operations, such as freezing accounts, suspending transactions, etc. When the risk level exceeds the threshold, the corresponding anti-fraud measures are automatically triggered through the smart contract: Trigger condition: Check whether the risk score of the node exceeds the threshold. If it exceeds, the smart contract automatically triggers the corresponding anti-fraud measures: Freeze the account: Freeze the high-risk account associated with this transaction.

[0118] Suspend the transaction: Suspend the transaction activities associated with this node.

[0119] Send a notice: Inform the relevant regulatory agencies or responsible parties through the smart contract or other notification mechanisms.

[0120] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent risk control system based on blockchain, characterized in that: include: Data integration module, used to integrate multiple data sources in real time through API interfaces or big data integration tools; A graph neural network model building module is used to convert the integrated data source into graph structure data, wherein: each node represents a participant, and the transaction relationship or upstream and downstream supply chain relationship between the nodes is used as an edge; each node carries feature data related to it; the graph neural network model models the relationship between nodes through a graph convolution operation to obtain a graph neural network model, thereby performing deep learning and analysis on the dependency relationship between nodes and their adjacent nodes in the graph; the graph convolution operation includes a multi-level convolution process, wherein: in the first layer of convolution, each node updates its own features by aggregating the information of its directly adjacent nodes; in the second and higher layers of convolution, the node aggregates the information of the adjacent nodes of its adjacent nodes to expand the scope of information propagation layer by layer, capture deeper dependencies, and help the model identify potential indirect effects; the multi-level convolution process helps the graph neural network gradually build a deep dependency network and form a global information aggregation between nodes, supports remote association analysis of nodes, and identifies fraudulent behavior propagated through indirect relationships; The fraud propagation chain identification module is used to identify the core nodes related to the fraud in the graph and trace the propagation path of the fraud. It iteratively identifies the directly and indirectly affected nodes through graph convolution operations, and updates the risk level of each node based on its influence value and position in the propagation chain. It outputs the complete fraud propagation path, including core nodes, indirect nodes, their risk levels and influence. A node feature updating module, used to update the features of each node according to the fraudulent behavior and propagation path identified by the fraudulent behavior propagation chain identification module; Real-time anti-fraud detection and response module, which is used to acquire and process transaction data in real time through streaming data processing technology, evaluate the fraud risk of each transaction node, and trigger the early warning and response mechanism when potential fraud risks are identified. The processing of transaction data is recorded and audited through blockchain technology; The fraud warning module is used to evaluate the risk level of each transaction node in real time based on the results of the node feature update module in combination with a machine learning algorithm, and take measures through smart contracts when potential fraud risks are identified. The smart contracts are executed through blockchain technology.

2. According to the blockchain-based intelligent risk control system of claim 1, it is characterized in that: The graph neural network model building module updates the features of each node in the following way: In the graph neural network model, the features of each node are iteratively calculated through graph convolution operations. The specific update process is as follows: in, is the feature representation of node v in the k+1th round, is the weight matrix, is the set of adjacent nodes of node v, is the inter-node normalization coefficient, is an activation function; the node features are dynamically updated in combination with the information of adjacent nodes to capture the dependencies between nodes.

3. According to the blockchain-based intelligent risk control system of claim 1, it is characterized in that: The graph neural network model is used to identify the fraudulent behavior of the core nodes in the direct fraud chain, further comprising: identifying the core nodes related to the fraudulent behavior in the graph through the graph neural network model; determining the core nodes as the implementation nodes of the fraud through feature analysis of the core nodes; tracing the fraudulent behavior of the core nodes based on the graph neural network model, and finding their connection relationship with other nodes in the graph; tracing back the nodes at each level through graph convolution operations, and propagating the fraudulent behavior of the core nodes to the direct or indirect nodes associated with them; using the graph neural network model to identify indirect nodes, that is, nodes affected by intermediary nodes; iteratively updating the influence value of each node through graph convolution calculation, and finally identifying all affected indirect nodes; for each node, calculating its influence in the fraudulent behavior propagation chain; nodes with greater influence are considered to be core fraud nodes, and nodes with less influence are considered to be indirectly affected nodes; after calculating the influence, using the influence value of each node as an indicator of the node characteristics, and updating its risk level according to its position in the fraud chain; based on the calculation of the graph neural network, outputting the complete fraud propagation path, including core nodes, indirect nodes, and the risk level and influence of each node.

4. According to the blockchain-based intelligent risk control system of claim 1, it is characterized in that: The fraud propagation chain identification module is used to trace the propagation path of fraud in the supply chain. Through the back propagation algorithm, the propagation path starting from the core node is traced. The graph neural network model is used to identify potential affected nodes through feature propagation between nodes. For each node, its influence with the core node is calculated, and it is judged whether it is part of the fraud chain based on the size of the influence. The influence of each node is calculated by the following formula: in, For Node To Node The propagation coefficient between For Node Features at layer I.

5. According to the blockchain-based intelligent risk control system of claim 1, it is characterized in that: The node feature update module updates the features of each node according to the fraudulent behavior and propagation path identified by the fraudulent behavior propagation chain identification module, specifically including: propagating the information of the adjacent nodes to the target node through iterative calculation of the graph neural network model, gradually updating the feature data of the target node to reflect the impact of the fraudulent behavior; dynamically adjusting the risk level of the node based on the result of the node feature update, the node with greater impact is considered to be the core fraud node, and the node with less impact is considered to be the indirectly affected node; the node feature update process updates the features of the node and its adjacent nodes based on the graph convolution operation; The risk level of the node is dynamically adjusted through multiple rounds of iterative calculations of the graph neural network, and the node features are gradually updated. After each round of iteration, the risk level of the node is adjusted according to its feature update; For each node, its risk score is calculated based on its distance from the fraudulent node or the influence spread in the graph neural network model, and the risk level is dynamically updated; The risk level update formula of a node is as follows: in, is the risk level of node v in round k+1, is the influence of node v, is the update function of the risk score.

6. According to the blockchain-based intelligent risk control system of claim 1, it is characterized in that: The node feature update module further expands the propagation range of node features through iterative calculations of the graph convolutional network, specifically including: in the process of node feature update, the feature information of adjacent nodes is gradually transmitted to the target node, and expanded to more distant nodes through multiple rounds of iterations to ensure that the impact of fraudulent behavior can be spread throughout the supply chain network. After each round of iteration, the node features are updated, and the node features ultimately reflect the potential risks of the node in the fraud chain.

7. According to claim 1, a blockchain-based intelligent risk control system is characterized in that: The real-time anti-fraud detection and response module monitors and responds to potential fraudulent behaviors in real time in the following ways: real-time monitoring of transaction data, obtaining transaction information through streaming data processing technology, and triggering the anti-fraud response mechanism when high-risk transactions are detected; real-time evaluation of the fraud risk of each transaction node through streaming computing, and triggering automatic account freezing and transaction suspension measures when necessary.

8. According to claim 1, a blockchain-based intelligent risk control system is characterized in that: The real-time anti-fraud detection and response module evaluates the fraud risk of each transaction node by combining a machine learning algorithm. The specific evaluation method is as follows: the feature data of each node is trained using a machine learning algorithm, and the transaction node is evaluated for real-time fraud risk using the trained model. When the risk value of the node exceeds a preset threshold, the corresponding anti-fraud measures are triggered.

9. According to the blockchain-based intelligent risk control system of claim 1, it is characterized in that: The fraud warning module automatically executes anti-fraud measures through smart contracts, specifically including: based on the risk level of the node, automatically executing anti-fraud operations through smart contracts. When the node risk level exceeds the threshold, the smart contract automatically executes measures such as freezing the account, suspending transactions, and notifying relevant parties; all anti-fraud measures are executed through smart contracts.

10. The blockchain-based intelligent risk control system according to claim 1, characterized in that: The multiple data sources include transaction data, supply chain data, and external data sources, and the data sources include transaction records, payment history, credit reports, and account change data of each node; Each transaction data is encrypted using the hash algorithm SHA-256 to generate a unique hash value for each transaction; the hash value is stored through the blockchain.

Images