Financial transaction verification method and system based on block chain and artificial intelligence

Abstract

The invention relates to a financial transaction verification method and system based on a block chain and artificial intelligence, and the method comprises the steps: storing financial transaction data through a multi-level block chain network, carrying out the feature extraction and model training of the preprocessed data through an artificial intelligence large model, verifying a new financial transaction, and outputting a result. The method comprises the following specific steps: storing financial transaction data through a multi-level block chain network; performing preprocessing operation on the data in the block chain network; performing feature extraction on the preprocessed data by using an artificial intelligence large model to form a feature vector; taking the feature vectors as input data, and performing training and optimization by using an artificial intelligence large model; verifying the new financial transaction according to the steps and outputting a result; updating and optimizing the model through a feedback mechanism; according to the invention, by combining the block chain and the artificial intelligence technology, the efficiency, accuracy and security of financial transaction verification are improved, and meanwhile, the privacy of data and the robustness of the model are ensured.

Description

Technical Field

[0001] The present invention relates to the field of fintech, and specifically to a financial transaction verification method and system based on blockchain and artificial intelligence. Background Art

[0002] With the rapid development of fintech, the frequency and complexity of financial transactions have increased significantly. Traditional financial transaction verification methods face many challenges. In the traditional financial system, transaction verification is usually completed by centralized institutions (such as banks). This method is not only inefficient but also has the risk of single-point failure and is difficult to meet the needs of modern financial transactions. The emergence of blockchain technology provides a new solution for financial transaction verification. Blockchain ensures that all nodes reach an agreement on the verification of transactions and blocks through a decentralized consensus mechanism, thus guaranteeing the authenticity and immutability of transactions. Common consensus mechanisms include Proof of Work (PoW), Proof of Stake (PoS), and Delegated Proof of Stake (DPoS). Among them, the DPoS mechanism has received extensive attention due to its high efficiency and low energy consumption. The DPoS mechanism elects representative nodes to participate in the consensus process of the blockchain network, reducing the centralization degree of the entire network and improving the transaction speed at the same time. However, the application of blockchain technology in financial transaction verification still faces some challenges. For example, the transaction verification process of blockchain needs to process a large amount of data, and the privacy and security of this data are important issues. In addition, the scalability and extensibility of blockchain also need to be further improved.

[0003] The development of artificial intelligence technology provides new tools for financial transaction verification. Artificial intelligence large models can extract key features in transaction data through learning and analysis of a large amount of data, thereby improving the accuracy and efficiency of transaction verification. For example, deep graph network technology can use convolutional neural networks to extract the features of text and transaction graphs, while the Transformer model can learn the correlation relationships in serialized data. However, the application of artificial intelligence technology in financial transaction verification also faces some challenges. For example, the training of artificial intelligence models requires a large amount of data, and the privacy and security of this data are important issues. In addition, the interpretability and robustness of artificial intelligence models also need to be further improved. Summary of the Invention

[0004] The main object of the present invention is to provide a financial transaction verification method and system based on blockchain and artificial intelligence, aiming to solve the limitations of traditional financial transaction verification methods and improve the efficiency, accuracy, and security of transaction verification.

[0005] A financial transaction verification method based on blockchain and artificial intelligence, characterized by including the following steps: S1. Store financial transaction data through a multi-level blockchain network; S2. Perform preprocessing operations on the data in the blockchain network; S3. Use an artificial intelligence large model to extract features from the preprocessed data to form feature vectors; S4. Use the feature vectors as input data and use the artificial intelligence large model for training and optimization; S5. Verify new financial transactions according to the above steps and output results; S6. Update and optimize the model through a feedback mechanism.

[0006] Further, the multi-level blockchain network includes a main chain and a side chain; the main chain adopts an improved DPoS consensus mechanism, which is used to store core transaction data and dynamically desensitize according to data types. The dynamic desensitization automatically switches the desensitization level according to the transaction amount threshold. Small transactions retain the hash ID, and large transactions add encrypted shards of the regulatory agency's key; the side chain deploys a federated learning model, which is used to store feature data and artificial intelligence large model parameters, and interacts with the main chain through a zero-knowledge proof protocol.

[0007] Further, the verification nodes of the main chain need to hold digital certificates issued by financial regulatory agencies; a privacy data access control protocol is designed inside the side chain, and different data processing permissions are assigned according to member identities.

[0008] Further, the preprocessing operations include performing consistency verification and distributed normalization processing on the data of the main chain and the side chain; using zk-STARK proof to verify the data consistency between the main chain and the side chain, encrypting sensitive fields with the SM4 national cryptography algorithm, and retaining plaintext indexes for non-sensitive fields.

[0009] Further, the side chain uses a distributed storage method to classify and store feature data, isolate user information, and establish indexes to provide fast retrieval services for generating feature vectors.

[0010] Further, the feature extraction includes: introducing deep graph network technology, using a convolutional neural network to extract the features of text and transaction graphs, and using a Transformer model to learn the correlation relationships in serialized data.

[0011] Further, the federated learning model uses a gradient transfer protocol to update model parameters and introduces a dynamic feature selection mechanism to filter out abnormal attack features; the training using the artificial intelligence large model includes encrypting the training samples of each institution with a homomorphic encryption algorithm.

[0012] Further, the verification of the new financial transaction according to the above steps and the output of the result include: allocating a verification channel based on the transaction complexity, calling the cross-chain credit investigation interface data for complex transactions, verifying the legality of the data source through the zero-knowledge proof protocol by the cross-chain interface, and mining the transaction trajectory in the money laundering scenario based on the community discovery algorithm and time series analysis.

[0013] Further, the update and optimization of the model through the feedback mechanism include inputting the intelligent decision verification result into the feedback system, aggregating and analyzing historical transactions through the blockchain smart contract, optimizing the rule base, and reducing the misjudgment rate; inputting the model with unqualified training sample quality into the adversarial sample generator to enhance the feature recognition ability; adopting the rolling release method to update the model to ensure unified learning of the model among various institutions and reduce the response time of the business to the model; obtaining sample and environment information from transaction feedback and model learning to form a positive feedback loop.

[0014] This application also proposes a financial transaction verification system based on blockchain and artificial intelligence, including the following modules: Blockchain network module: storing financial transaction data through a multi-level blockchain network; Data processing module: performing preprocessing operations on the data in the blockchain network; Feature extraction module: using an artificial intelligence large model to extract features from the preprocessed data to form feature vectors; Model training and optimization module: using the feature vectors as input data and training and optimizing them with an artificial intelligence large model; Transaction verification module: using the feature vectors as input data and training and optimizing them with an artificial intelligence large model; Feedback and optimization module: updating and optimizing the model through a feedback mechanism.

[0015] A financial transaction verification method and system based on blockchain and artificial intelligence provided by the present invention stores financial transaction data through a multi-level blockchain network. By utilizing the immutable and distributed storage characteristics of the blockchain, it can effectively prevent data from being maliciously tampered with, while ensuring the transparency and traceability of the data, providing a secure and reliable foundation for financial transaction verification. The introduction of the artificial intelligence large model enables the system to quickly extract features and identify patterns from a large amount of financial transaction data. Compared with traditional rule-based verification methods, it can more efficiently and accurately judge the authenticity and legality of transactions, reducing the workload of manual review and potential human errors. The flexibility and scalability of feature extraction and model training enable it to cope with constantly changing financial transaction scenarios and patterns. Whether it is a simple personal transfer transaction or a complex cross-border financial transaction, the system can provide targeted verification services through gradually optimized features and models, effectively identifying various financial risks. The feedback-driven closed-loop optimization mechanism realizes the continuous update and improvement of the model, ensuring that the system can promptly adapt to the changes in new financial transaction rules and regulatory requirements, continuously enhancing its own performance, and ensuring a high verification accuracy and reliability in practical applications. While providing strong support for financial institutions and regulatory authorities, it also helps to maintain the stability and security of the financial market. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a schematic flowchart of a financial transaction verification method based on blockchain and artificial intelligence according to an embodiment of the present application.

[0017] The implementation, functional characteristics, and advantages of the object of the present invention will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] In order to make the object, technical solution, and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0019] Those skilled in the art of the present technology can understand that unless otherwise defined, all terms (including technical terms and scientific terms) used herein have the same meaning as generally understood by those of ordinary skill in the art to which the present invention belongs. It should also be understood that terms such as those defined in a general dictionary should be understood to have a meaning consistent with the meaning in the context of the prior art, and will not be interpreted with an idealized or overly formal meaning unless specifically defined as here.

[0020] Referring to Figure 1 , in order to achieve the above object of the invention, the present invention provides a financial transaction verification method based on blockchain and artificial intelligence, including the following steps: A financial transaction verification method based on blockchain and artificial intelligence, characterized by including the following steps: S1. Store financial transaction data through a multi-level blockchain network; S2. Perform preprocessing operations on the data in the blockchain network; S3. Use an artificial intelligence large model to extract features from the preprocessed data to form feature vectors; S4. Use the artificial intelligence large model to train and optimize with the feature vectors as input data; S5. Verify new financial transactions according to the above steps and output results; S6. Update and optimize the model through a feedback mechanism.

[0021] In this embodiment, financial transaction data is stored through a multi-level blockchain network. When a financial transaction occurs, the relevant data will be digitized and uploaded to the blockchain network. A multi-level blockchain architecture is used here, which can better adapt to the needs of large-scale data storage and high concurrent access. Through the distributed storage and consensus mechanism of the blockchain, the data is guaranteed to be tamper-proof and traceable, and a reliable data source is provided for subsequent data processing and analysis. The data in the blockchain network is preprocessed. The data stored in the blockchain network needs to be preprocessed before entering the subsequent processing steps. This process is mainly to clean up the noise, incomplete or erroneous data in the data. At the same time, the data is normalized to meet the requirements of subsequent feature extraction and model training. Preprocessing is an important part of data analysis and machine learning. It can improve data quality and reduce the interference of abnormal data on model training, thereby improving the performance and accuracy of the entire system; use artificial intelligence big models to extract features from preprocessed data to form feature vectors. Feature extraction is one of the key steps in machine learning. At this stage, the system will use artificial intelligence big models to conduct in-depth analysis of the preprocessed data and explore the potential features contained in the data. These features can be the changing trend of transaction amount, the law of transaction frequency, the relationship between the two parties to the transaction, etc., which are highly related to financial transaction verification. The powerful computing power and ability to understand complex patterns of the big model enable it to extract rich and high-quality features from large amounts of data, providing strong support for subsequent model training and transaction verification; the feature vector is used as input data, and the artificial intelligence big model is used for training and optimization. After the feature extraction is completed, the obtained feature vector will be used as input data to train the artificial intelligence big model. This model will learn based on the verification results in the historical transaction data, and continuously adjust the model's parameters and structure through the deep learning algorithm to improve its ability to judge the authenticity and legality of the transaction. The model training process is an iterative optimization process. Through continuous trial and error and adjustment, the model can gradually Gradually grasp the patterns and rules in transaction data, so that when faced with new transaction data, more accurate predictions and judgments can be made; new financial transactions are verified according to the above steps and the results are output. When a new financial transaction occurs, the system will process the transaction data according to the previous steps, including data storage, preprocessing, feature extraction and model prediction; the trained artificial intelligence large model will analyze the new transaction data, and based on the learned features and patterns, judge the authenticity and legality of the transaction, and finally output the verification results. This result can be a simple binary judgment (such as "legal" or "suspicious"), or a more detailed analysis report for reference by financial institutions and relevant regulatory authorities;The model is updated and optimized through a feedback mechanism. Since the environment and patterns of financial transactions are constantly changing, the model needs to continuously learn new data to keep up with market changes. The feedback mechanism allows the system to re-enter new transaction data and verification results into the model as new training samples, helping the model update its internal parameters and feature understanding, thereby improving the accuracy and adaptability of the model. This closed-loop optimization mechanism enables the system to continuously progress and adapt to new financial transaction environments and requirements.

[0022] In one embodiment, the multi-level blockchain network includes a main chain and a side chain; the main chain adopts an improved DPoS consensus mechanism for storing core transaction data and dynamically desensitizing according to data types. The dynamic desensitization automatically switches the desensitization level according to the transaction amount threshold. Small transactions retain the hash ID, and large transactions add encrypted regulatory agency key shards; the side chain deploys a federated learning model for storing feature data and artificial intelligence large model parameters and interacting with the main chain through a zero-knowledge proof protocol.

[0023] In this embodiment, the multi-level blockchain network is designed to include two parts: the main chain and the side chain, in order to achieve efficient data storage, secure management, and cross-chain interaction. The main chain adopts an improved DPoS (Delegated Proof of Stake) consensus mechanism, which is a consensus algorithm based on delegated proof of stake. Its core lies in electing representative nodes to proxy the entire network for transaction verification and block generation. Compared with traditional consensus mechanisms, the DPoS mechanism can significantly improve the transaction processing speed of the network while reducing energy consumption, which is particularly important for financial scenarios that require frequent and large-scale processing of transaction data. The main chain is mainly used to store the core data of transactions. These data have undergone elaborate dynamic data masking processing. The dynamic data masking technology can automatically adjust the data masking level according to the size of the transaction amount. For small transactions, only the hash ID is retained for traceability and basic security protection. For large transactions, the key sharding encryption of the regulatory agency will be added to form multi-layer protection. This not only enhances the security of the data but also ensures the reasonable utilization of the data in different transaction scenarios, meeting the requirements of financial transactions for data confidentiality and compliance. On the other hand, the side chain is deployed to store federated learning models, as well as feature data and related artificial intelligence large model parameters. Federated learning is an emerging machine learning technology that allows different institutions to collaboratively train models without sharing raw data, thus achieving knowledge sharing while protecting data privacy. Through the gradient transfer protocol, the federated learning framework can achieve distributed updates of model parameters, enabling the side chain to make full use of the data resources of each node for model optimization and training. At the same time, the side chain also interacts with the main chain through the zero-knowledge proof protocol. This protocol allows one party to prove the truth of a certain statement to the other party without disclosing any useful information, thus achieving data consistency verification between the main chain and the side chain while ensuring data privacy, and promoting the secure sharing and interaction of data on different chains.

[0024] In one embodiment, the main chain verification nodes need to hold digital certificates issued by financial regulatory agencies; a privacy data access control protocol is designed within the side chain, and different data processing permissions are assigned according to the membership status.

[0025] In this embodiment, in the blockchain network, the main-chain verification nodes are responsible for verifying the authenticity and legality of transactions. To ensure the credibility and authority of these nodes, they are required to hold digital certificates issued by financial regulatory authorities. A digital certificate is an electronic credential used to prove the identity and legality of a node, preventing malicious nodes from participating in the transaction verification process. This mechanism enhances the security and credibility of the main chain through identity authentication; this not only prevents the participation of malicious nodes but also enhances the security and credibility of the entire system, making financial institutions and users more confident in conducting transactions; the side chain undertakes the task of storing and processing specific data in the blockchain network. To protect the privacy and security of this data, a privacy data access control protocol is designed inside the side chain; this protocol assigns different data processing permissions according to the identities of members, ensuring that only authorized nodes can access and process sensitive data. This mechanism prevents the abuse and leakage of data through permission management, protecting the privacy of users. This mechanism ensures the privacy and security of data, meets the strict requirements for data protection in financial transactions, and enhances users' trust in the system.

[0026] In one embodiment, the preprocessing operation includes performing consistency verification and distributed normalization processing on the main-chain and side-chain data; using zk-STARK proofs to verify the consistency of the main-chain and side-chain data, encrypting sensitive fields using the SM4 national cryptography algorithm, and retaining plaintext indexes for non-sensitive fields.

[0027] In this embodiment, the data consistency between the main chain and the side chain is the key to ensuring the normal operation of the entire system. The zk-STARK (Zero-Knowledge Scalable Transparent ARgument of Knowledge) proof is used to verify the data consistency between the main chain and the side chain. zk-STARK is a zero-knowledge proof technology that allows one party to prove the authenticity of a certain statement to another party without disclosing any useful information. Specifically, zk-STARK constructs a zero-knowledge proof system, enabling the verifier to verify the data consistency and integrity without obtaining the specific data content. The application of this technology ensures the data synchronization and consistency between the main chain and the side chain, preventing data tampering and inconsistency issues; Distributed normalization processing standardizes the data to make it consistent and comparable between different nodes and chains. The normalization processing includes operations such as scaling and translation of the data to make it conform to a specific distribution or range. In a distributed system, normalization processing can improve the efficiency and accuracy of data processing, ensuring that the data between different nodes has a consistent format and scale, thus facilitating subsequent analysis and processing; The SM4 national cryptography algorithm is used to encrypt sensitive fields. Among them, SM4 is a symmetric cryptography algorithm released by the China National Cryptography Administration for data encryption and decryption. The key length of the SM4 algorithm is 128 bits, and the block length is also 128 bits, with high security and high performance. Encrypting sensitive fields with the SM4 national cryptography algorithm ensures the security of data during transmission and storage. The SM4 algorithm provides strong data protection capabilities through multiple rounds of encryption and complex transformation operations. At the same time, for non-sensitive fields, the plaintext index is retained for fast data retrieval and query. The plaintext index refers to directly using the original data as the index without encryption. This design improves the efficiency of data retrieval while ensuring data security, enabling the system to quickly access and process relevant data when needed.

[0028] In one embodiment, the side chain classifies and stores the feature data in a distributed storage manner, isolates the user information, and establishes an index to provide fast retrieval services for feature vector generation.

[0029] In this embodiment, the side chain uses a distributed storage method to classify and store feature data. Distributed storage means that data is dispersed and stored on multiple nodes, and each node is responsible for storing a part of the data. This storage method can improve the availability and fault tolerance of data, while reducing the risk of single-point failures. The feature data is dispersed and stored on multiple nodes, and data storage and retrieval tasks can be processed in parallel, thus improving the overall performance of the system. In the side chain, the feature data is classified and stored, that is, data of different types or from different sources are stored in different nodes or storage areas respectively for easy management and retrieval. The user information is isolated, that is, the feature data of different users are stored in different storage areas or nodes respectively. This can ensure the independence and security of user data, prevent the data of different users from interfering with each other or being leaked. The isolation of user information ensures the independence of the feature data of different users and prevents data leakage and cross-contamination. The isolation of user information is an important measure to protect user privacy and data security. To provide a fast retrieval service for feature vector generation, an index needs to be established. An index is a data structure used to improve the speed of data retrieval. In the side chain, by establishing an index for the feature data, the required feature vectors can be quickly located and retrieved, thus improving the response speed and efficiency of the system. Common index structures include inverted indexes, tree structures (such as KD-trees, B-trees), etc. Establishing an index can significantly improve the retrieval speed of feature vectors, especially when dealing with large-scale data sets. A fast retrieval service can shorten the response time of the system and improve the user experience. Distributed storage and index technology enable the system to efficiently process large-scale feature data and meet the high requirements for data processing in financial transaction verification. This plays an important role in improving the scalability and adaptability of the system.

[0030] In one embodiment, the feature extraction includes: introducing deep graph network technology, using a convolutional neural network to extract the features of text and transaction graphs, and using a Transformer model to learn the correlation relationships in the serialized data.

[0031] In this embodiment, the deep graph network technology is introduced. The deep graph network technology is a deep learning method for processing graph-structured data. By applying convolutional operations on the graph structure, it can effectively extract the features of nodes and their neighborhoods in the graph. In financial transaction verification, transactions can be regarded as nodes in the graph, and the relationships between transactions can be regarded as edges. Through the deep graph network technology, complex relationships and patterns in the transaction graph can be captured. The convolutional neural network is used to extract the features of the text and the transaction graph. The convolutional neural network is a deep learning model specifically designed for processing data with grid structures (such as images and text). The convolutional neural network (CNN) gradually extracts the local and global features of the data through convolutional layers and pooling layers. In financial transaction verification, the convolutional neural network (CNN) can be used to extract the features of transaction texts and transaction graphs, such as transaction descriptions, transaction amounts, etc. By introducing the deep graph network technology and the convolutional neural network, the features in the transaction graph and text can be extracted more effectively, improving the accuracy of feature extraction. At the same time, the Transformer model is used to learn the correlation relationships in the serialized data. The Transformer model is a deep learning model based on the self-attention mechanism, originally used for natural language processing tasks. It can effectively process serialized data and capture the long-range dependence relationships between elements in the sequence. In financial transaction verification, the Transformer model can be used to learn the time-series features and correlation relationships in the transaction sequence. The combination of the deep graph network and the Transformer model enables the model to better capture the complex patterns and long-range dependence relationships in the data, enhancing the model's expressive ability and generalization ability. This helps to improve the accuracy and reliability of financial transaction verification. At the same time, the combination of the deep graph network and the Transformer model has good performance when dealing with large-scale data, and can support the processing requirements of large-scale data in financial transaction verification. This helps to improve the scalability and adaptability of the system.

[0032] In one embodiment, the federated learning model implements model parameter updates using a gradient transfer protocol and introduces a dynamic feature selection mechanism to filter out abnormal attack features; the training using the artificial intelligence large model includes encrypting the training samples of each institution using a homomorphic encryption algorithm.

[0033] In this embodiment, the federated learning model uses a gradient transmission protocol to update model parameters. The gradient transmission protocol is a communication protocol in federated learning, which is used to transmit the gradient information of the model among participants, so as to update the model parameters. Specifically, each participant calculates the gradient of the model locally, and then encrypts and sends these gradients to the central server. The central server aggregates these gradients and updates the global model parameters, and then sends the updated parameters back to each participant. This process ensures the privacy of data and the collaborative training effect of the model. The dynamic feature selection mechanism is a method of dynamically selecting and adjusting features during model training. By analyzing the importance of features, the dynamic feature selection mechanism can filter out those features that are not helpful for model training or may even be exploited by abnormal attackers. Specifically, statistical methods (such as SHAP values) can be used to evaluate the contribution of each feature to the model prediction result, and then the feature set can be dynamically adjusted according to these evaluation results, so as to improve the robustness and accuracy of the model. At the same time, the training samples of each institution are encrypted using the homomorphic encryption algorithm. Homomorphic encryption is an encryption technology that allows direct calculation on encrypted data without decrypting the data first. In federated learning, the training samples of each institution can be encrypted using the homomorphic encryption algorithm and then the model training can be carried out in the encrypted state. Specifically, the homomorphic encryption algorithm allows addition and multiplication operations on ciphertext data, thus supporting the model training process. This ensures the privacy and security of data during transmission and processing. Even if the data is intercepted during transmission, the attacker cannot obtain the plaintext data, thus effectively preventing data leakage.

[0034] In one embodiment, the verification of the new financial transaction according to the above steps and the output of the result include: allocating a verification channel based on the transaction complexity, and complex transactions call the data of the cross-chain credit investigation interface. The cross-chain interface verifies the legality of the data source through the zero-knowledge proof protocol, and mines the transaction track in the money laundering scenario based on the community discovery algorithm and time series analysis.

[0035] In this embodiment, verification channels are allocated based on transaction complexity. Specifically, simple transactions are subject to basic verification through a lightweight model, while complex transactions call data from the cross-chain credit investigation interface. This allocation mechanism can effectively improve verification efficiency and ensure that complex transactions are more comprehensively and deeply verified. For complex transactions, the system will call data from the cross-chain credit investigation interface. Cross-chain technology allows data interaction between multiple blockchain networks, and the credit investigation interface provides access to external credit data. Through the cross-chain credit investigation interface, the system can obtain more comprehensive credit information, thereby improving the accuracy of transaction verification. The cross-chain interface verifies the legality of the data source through the zero-knowledge proof protocol. The zero-knowledge proof protocol is a cryptographic technology that allows one party to prove the truth of a statement to another party without revealing any useful information. The cross-chain interface verifies the legality of the data source through the zero-knowledge proof protocol, ensuring the authenticity and reliability of the data while protecting the privacy of the data. Based on the community discovery algorithm and time series analysis, the transaction trajectories in money laundering scenarios are mined. The community discovery algorithm is used to identify the community structure in the transaction network. By analyzing the network topology structure of transaction data, potential fraud communities or money laundering gangs can be discovered. Time series analysis is used to analyze the time series characteristics of transaction data and mine the rules and abnormal patterns of transaction behavior. Combining these two algorithms, the system can more accurately identify the transaction trajectories in money laundering scenarios and improve the accuracy of money laundering detection.

[0036] In one embodiment, the update and optimization of the model through the feedback mechanism include inputting the intelligent decision verification results into the feedback system, aggregating and analyzing historical transactions through blockchain smart contracts to optimize the rule base and reduce the misjudgment rate; inputting the model with unqualified training sample quality into the adversarial sample generator to enhance the feature recognition ability; using the rolling release method to update the model to ensure unified learning of the model among various institutions and reduce the response time of the business to the model; obtaining sample and environmental information from transaction feedback and model learning to form a positive feedback loop.

[0037] In this embodiment, the intelligent decision verification results are input into the feedback system, and these results are used to adjust and improve the model. This feedback mechanism enables the model to continuously learn and adapt according to the actual verification results, thereby improving the accuracy and robustness of the model. Through the aggregation and analysis of historical transactions by blockchain smart contracts, which are self-executing contract terms that can ensure data transparency and immutability, the system can automatically collect and analyze historical transaction data, optimize the rule base, and reduce the misjudgment rate. This mechanism not only improves the efficiency of data processing but also ensures the accuracy and security of the data. For models with substandard training sample quality, they are input into the adversarial sample generator, which generates challenging samples through technologies such as generative adversarial networks (GANs) to enhance the model's feature recognition ability. This mechanism can improve the model's ability to recognize abnormal and attack features and enhance the model's robustness. The rolling release method is used to update the model to ensure unified learning of the model among various institutions and reduce the response time of the business to the model. Rolling release is a deployment strategy that gradually replaces old version instances and can complete the model update without interrupting the service. This method can reduce the impact on users and ensure the continuous optimization of the model. Through transaction feedback and model learning, the system can obtain more sample and environmental information, forming a positive feedback loop. This mechanism enables the model to continuously learn and adapt to new data and environments, further improving the performance and accuracy of the model.

[0038] This application also proposes a financial transaction verification system based on blockchain and artificial intelligence, including the following modules: Blockchain network module: storing financial transaction data through a multi-level blockchain network; Data processing module: performing preprocessing operations on the data in the blockchain network; Feature extraction module: using an artificial intelligence large model to extract features from the preprocessed data to form feature vectors; Model training and optimization module: using the feature vectors as input data and training and optimizing them using an artificial intelligence large model; Transaction verification module: using the feature vectors as input data and training and optimizing them using an artificial intelligence large model; Feedback and optimization module: updating and optimizing the model through a feedback mechanism.

[0039] The operation mode of the device in this embodiment refers to the foregoing method embodiment and will not be elaborated here.

[0040] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium provided in the present application and used in the embodiments can include non-volatile and / or volatile memories. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, there are various forms of RAM, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (RambuS) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0041] In summary, a financial transaction verification method and system based on blockchain and artificial intelligence provided by the present invention stores financial transaction data through a multi-level blockchain network. By utilizing the immutable and distributed storage characteristics of the blockchain, it can effectively prevent data from being maliciously tampered with, while ensuring data transparency and traceability, providing a secure and reliable foundation for financial transaction verification. The introduction of the artificial intelligence large model enables the system to quickly extract features and identify patterns from a large amount of financial transaction data. Compared with traditional rule-based verification methods, it can more efficiently and accurately judge the authenticity and legality of transactions, reducing the workload of manual review and potential human errors. The flexibility and scalability of feature extraction and model training enable it to cope with constantly changing financial transaction scenarios and patterns. Whether it is a simple personal transfer transaction or a complex cross-border financial transaction, the system can provide targeted verification services through gradually optimized features and models, effectively identifying various financial risks. The feedback-driven closed-loop optimization mechanism realizes the continuous update and improvement of the model, ensuring that the system can promptly adapt to new financial transaction rules and changes in regulatory requirements, continuously enhancing its own performance, and ensuring a high verification accuracy and reliability in practical applications. While providing strong support for financial institutions and regulatory authorities, it also helps to maintain the stability and security of the financial market.

[0042] The above are only the preferred embodiments of the present application, and do not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made by using the content of the specification and drawings of the present application, or directly or indirectly applied in other related technical fields, shall be equally included in the patent protection scope of the present application.

Claims

1. A financial transaction verification method based on blockchain and artificial intelligence, characterized in that: The following steps are involved: S1. Store financial transaction data through a multi-level blockchain network; S2. Preprocess the data in the blockchain network; S3. Use the artificial intelligence big model to extract features from the preprocessed data to form a feature vector; S4, using the feature vector as input data, and using the artificial intelligence big model for training and optimization; S5. Verify the new financial transaction according to the above steps and output the result; S6. Update and optimize the model through feedback mechanism.

2. The financial transaction verification method according to claim 1, characterized in that: The multi-level blockchain network includes a main chain and a side chain; The main chain adopts an improved DPoS consensus mechanism to store core transaction data and dynamically desensitizes data according to the data type. The dynamic desensitization automatically switches the desensitization level according to the transaction amount threshold. Small transactions retain hash IDs, and large transactions add regulatory agency key sharding encryption; The side chain deploys a federated learning model to store feature data and artificial intelligence model parameters, and interacts with the main chain through a zero-knowledge proof protocol.

3. The financial transaction verification method according to claim 2, characterized in that: The main chain verification node must hold a digital certificate issued by the financial regulatory authority; a privacy data access control protocol is designed in the side chain to assign different data processing permissions based on member identity.

4. The financial transaction verification method according to claim 2, characterized in that: The preprocessing operation includes consistency verification and distributed normalization of the main chain and side chain data; using zk-STARK proof to verify the consistency of the main chain and side chain data, encrypting sensitive fields with the SM4 national secret algorithm, and retaining plaintext indexes for non-sensitive fields.

5. The financial transaction verification method according to claim 2, characterized in that: The side chain uses a distributed storage method to classify and store feature data, isolate user information, and create an index to provide fast retrieval services for feature vector generation.

6. The financial transaction verification method according to claim 1, characterized in that: The feature extraction includes: introducing deep graph network technology, using convolutional neural network to extract features of text and transaction graphs, and using Transformer model to learn the association relationship in serialized data.

7. The financial transaction verification method according to claim 2, characterized in that: The federated learning model adopts a gradient transfer protocol to implement model parameter updates, and introduces a dynamic feature selection mechanism to filter abnormal attack features; The use of a large artificial intelligence model for training includes encrypting the training samples of each institution using a homomorphic encryption algorithm.

8. The financial transaction verification method according to claim 1, characterized in that: The verification of new financial transactions according to the above steps and output of results include: allocating verification channels based on transaction complexity, complex transactions calling cross-chain credit interface data, the cross-chain interface verifies the legitimacy of the data source through a zero-knowledge proof protocol, and mining transaction tracks in money laundering scenarios based on community discovery algorithms and time series analysis.

9. The financial transaction verification method according to claim 1, characterized in that: The updating and optimization of the model through the feedback mechanism includes inputting the intelligent decision-making verification results into the feedback system, performing aggregate analysis on historical transactions through blockchain smart contracts, optimizing the rule base, and reducing the misjudgment rate; Input the model with substandard training samples into the adversarial sample generator to enhance feature recognition capabilities; Adopt a rolling release approach to update the model to ensure unified learning across organizations and reduce business response time to the model; Obtain sample and environmental information from transaction feedback and model learning to form a positive feedback closed loop.

10. A financial transaction verification system based on blockchain and artificial intelligence, including the following modules: Blockchain network module: store financial transaction data through a multi-level blockchain network; Data processing module: pre-processing the data in the blockchain network; Feature extraction module: Use the artificial intelligence big model to extract features from the preprocessed data to form a feature vector; Model training and optimization module: takes feature vectors as input data and uses large artificial intelligence models for training and optimization; Transaction verification module: takes feature vectors as input data and uses artificial intelligence big models for training and optimization; Feedback and optimization module: Update and optimize the model through feedback mechanism.

Images