Decentralized transportation credit management method and device based on block chain

Abstract

The invention relates to the technical field of block chains, in particular to a decentralized transportation credit management method and device based on a block chain. The method comprises the following steps: storing credit data of traffic participants into corresponding fragment chains in a multi-chain collaborative architecture in a classified manner; the fragment chain comprises a vehicle chain, a driver chain and an enterprise chain; performing bidirectional anchoring verification on the credit data hash value between the main chain and the fragment chain, generating a cross-chain verification certificate and synchronizing the cross-chain verification certificate to a federated learning module of the main chain; verifying the cross-chain verification certificate through a federated learning module, and performing encryption aggregation on the local training gradient of each fragment chain by adopting a secure multi-party computing technology according to cross-chain data under the condition that the verification is passed, so as to generate a unified dynamic credit assessment model; and under the condition that the real-time intelligent contract on the fragment chain is controlled to capture the latest traffic behavior data, calling the dynamic credit evaluation model to execute credit score calculation, and writing a score result into a new block of the corresponding fragment chain.

Description

Technical Field

[0001] The present application relates to the field of blockchain technology, and in particular to a decentralized transportation credit management method and device based on blockchain. Background Art

[0002] In the field of traffic management, accurate and efficient management of the credit data of traffic participants is crucial. With the continuous increase in the number of traffic participants and the increasing complexity of credit data types, traditional data storage and management methods are no longer able to meet the needs.

[0003] As an emerging distributed data storage and processing solution, the multi-chain collaborative architecture provides new ideas for the classified storage and management of credit data of transportation participants.

[0004] However, how to accurately and securely classify and store the credit data of different types of traffic participants in the corresponding shard chains in the multi-chain collaborative architecture, and ensure the integrity and security of the data, is an urgent problem that needs to be solved. Summary of the Invention

[0005] Based on this, it is necessary to address the above technical problems and provide a blockchain-based decentralized transportation credit management method and device that can classify and store the credit data of transportation participants in the corresponding shard chains in the multi-chain collaborative architecture.

[0006] In the first aspect, the present application provides a decentralized transportation credit management method based on blockchain, which includes: The credit data of traffic participants are classified and stored in the corresponding shard chains in the multi-chain collaborative architecture; the shard chains include vehicle chains, driver chains and enterprise chains; Perform two-way anchor verification on the credit data hash value between the main chain and the shard chain, generate a cross-chain verification certificate and synchronize it to the federated learning module of the main chain; The cross-chain verification certificate is verified through the federated learning module. If the verification passes, the local training gradients of each shard chain are encrypted and aggregated using secure multi-party computing technology based on the cross-chain data to generate a unified dynamic credit assessment model. The real-time smart contract on the control shard chain captures the latest traffic behavior data, calls the dynamic credit assessment model to perform credit score calculation, and writes the score result into a new block of the corresponding shard chain.

[0007] In one embodiment, the credit data of the transportation participants are classified and stored in the corresponding shard chains in the multi-chain collaborative architecture, including: Write vehicle violation data into the vehicle chain, write driver credit data into the driver chain, and write transportation enterprise data into the enterprise chain.

[0008] In one embodiment, the credit data of the transportation participants are classified and stored in the corresponding shard chains in the multi-chain collaborative architecture, including: Verify the credit data storage request and, after meeting the preset credit data integrity verification conditions, package the encrypted data into the block of the corresponding shard chain; Among them, the encrypted data includes the timestamp data of the frequency of running red lights in the vehicle chain, the credit repair operation records in the driver chain, and the complaint response time log in the enterprise chain.

[0009] In one embodiment, based on cross-chain data, secure multi-party computing technology is used to encrypt and aggregate the local training gradients of each shard chain to generate a unified dynamic credit assessment model, including: Extract encrypted red light running frequency timestamp data from the vehicle chain and generate a violation feature vector including a time decay factor; Analyze the credit repair operation records from the driver chain and generate a behavioral feature vector containing the repair time weight; Analyze complaint response time logs from the enterprise chain and generate a service quality feature vector based on response speed; The violation feature vector, behavior feature vector, and service quality feature vector are input into the federated learning model, fused through the spatiotemporal weight matrix, and output as a dynamic credit assessment function. Based on the output dynamic credit assessment function, a unified dynamic credit assessment model is constructed.

[0010] In one embodiment, the real-time smart contract on the control shard chain captures the latest traffic behavior data and calls the dynamic credit assessment model to perform credit score calculation, including: When the real-time smart contract on the control shard chain receives a traffic behavior event trigger, it extracts the time decay factor and repair time weight in the dynamic credit evaluation function; Combined with the spatiotemporal attributes of the current event, encrypted scoring parameters are generated through homomorphic encryption calculations; Perform matrix operations on the encrypted scoring parameters and the dynamic credit assessment function, and output the credit score ciphertext containing a verifiable timestamp.

[0011] In one embodiment, after writing the scoring result into a new block of the corresponding shard chain, the method further includes: Based on the credit score deviation data monitored in real time by the shard chain, the main chain is triggered to adjust the local parameters of the dynamic credit assessment model through the parameter optimization algorithm of the federated learning model, and the updated parameters are synchronized to the relevant shard chains through the cross-chain protocol.

[0012] In one embodiment, after synchronizing the updated parameters to the relevant shard chains via the cross-chain protocol, the method further includes: The control shard chain performs retroactive calculations on historical encrypted scoring parameters based on the updated dynamic credit assessment function; According to the backtracking calculation results, the correction parameters are activated and the spatiotemporal dimension weight matrix of the federated learning model is updated.

[0013] In one embodiment, based on the backtracking calculation results, activating the correction parameters and updating the spatiotemporal weight moments of the federated learning model include: Perform differential comparison between the backtracking calculation results and the expected correction value of the main chain to generate a weight matrix correction verification report; When the difference value is less than the preset tolerance, the correction parameters are activated and the spatiotemporal dimension weight matrix of the federated learning model is updated.

[0014] In one embodiment, after activating the correction parameters and updating the spatiotemporal weight matrix of the federated learning model based on the backtracking calculation results, the method further includes: When the driver chain credit repair operation record is monitored, the main chain is triggered to update the repair time weight of the dynamic credit evaluation function; Send weight update instructions to the vehicle chain and enterprise chain through the cross-chain protocol to synchronously adjust the credit score parameters of the associated entities; The updated scoring parameters are input into the federated learning model to generate a new version of the dynamic credit assessment function.

[0015] Secondly, a decentralized transportation credit management device based on blockchain is provided, including: The data acquisition module is used to classify and store the credit data of traffic participants in the corresponding shard chains in the multi-chain collaborative architecture; the shard chains include vehicle chains, driver chains, and enterprise chains; The multi-center verification module is used to perform two-way anchor verification on the credit data hash value between the main chain and the shard chain, generate cross-chain verification certificates and synchronize them to the federated learning module of the main chain; The unified model training module is used to verify the cross-chain verification certificate through the federated learning module. If the verification is passed, the local training gradients of each shard chain are encrypted and aggregated using secure multi-party computing technology based on the cross-chain data to generate a unified dynamic credit assessment model; The credit scoring module is used to control the real-time smart contract on the shard chain to call the dynamic credit assessment model to perform credit scoring calculations while capturing the latest traffic behavior data, and write the scoring results into the new block of the corresponding shard chain.

[0016] The aforementioned blockchain-based decentralized transportation credit management method and device categorizes and stores the credit data of transportation participants in corresponding shard chains, such as the vehicle chain, driver chain, and enterprise chain, enabling refined management of credit data. Storing and processing different types of data on their respective shard chains not only facilitates rapid data retrieval and maintenance, but also enables targeted data processing and analysis based on the characteristics and needs of different participants, improving the efficiency and flexibility of data management.

[0017] A federated learning module verifies cross-chain verification certificates. Once verified, secure multi-party computing (SMC) technology is used to encrypt and aggregate the local training gradients of each shard chain to generate a unified dynamic credit assessment model. This model comprehensively considers various credit data from different shard chains and dynamically adjusts and optimizes based on real-time traffic behavior data. Compared to traditional static credit assessment models, it can more accurately reflect the true credit status of traffic participants.

[0018] The real-time smart contract on the control shard chain captures the latest traffic behavior data, invokes a dynamic credit assessment model to calculate a credit score, and writes the score result to a new block on the corresponding shard chain. This process enables real-time credit score updates, reflecting the latest performance of traffic participants. This provides timely and accurate credit information to traffic management departments, businesses, and other relevant parties, facilitating better traffic management and decision-making, such as rewarding drivers or businesses with high credit ratings and implementing appropriate supervision and constraints on those with low credit ratings. The decentralized blockchain-based architecture eliminates reliance on a single central authority, preventing potential single points of failure and data misuse. Furthermore, secure multi-party computation technology protects the privacy of data on each shard chain during the encrypted aggregation process. Data from all participants is jointly computed without leaking, ensuring both data availability and the privacy of data owners. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments of the present application or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying any creative work.

[0020] Figure 1 1. A flowchart of a decentralized transportation credit management method based on blockchain in one embodiment; Figure 2 Schematic diagram of a decentralized transportation credit management device based on blockchain in one embodiment. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0022] In an exemplary embodiment, Figure 1 As shown, a decentralized transportation credit management method based on blockchain is provided, which is applied to the blockchain. The blockchain includes a main chain (100), a shard chain (200), a dynamic credit evaluation model (300) and a smart contract (400). The method includes: S1. Classify and store the credit data of traffic participants in the corresponding shards in the multi-chain collaborative architecture.

[0023] Among them, the shard chain includes vehicle chain, driver chain and enterprise chain.

[0024] Optionally, a collaborative architecture of the main chain (100) and the shard chain (200) is constructed, and the shard chain includes: vehicle chain (210): stores vehicle violation data (such as license plate number, violation time, violation type); driver chain (220): stores driver credit data (such as driver's license deduction records, credit repair records); enterprise chain (230): stores transportation enterprise data (such as operating license number, complaint handling records).

[0025] Data storage rules: The vehicle chain uses spatiotemporal sharding technology, dividing data storage nodes by geographic location. The driver chain and enterprise chain use attribute-encrypted storage, and data access requires identity certificate verification. For example, when a vehicle (license plate number: Beijing A12345) runs a red light, the traffic police system will write a violation record containing the timestamp (2023-08-2014:30) and geographic location (longitude 116.40, latitude 39.90) to the corresponding geographic shard node of the vehicle chain.

[0026] S2. Perform two-way anchor verification on the credit data hash value between the main chain and the shard chain, generate a cross-chain verification certificate and synchronize it to the federated learning module of the main chain.

[0027] Cross-chain verification process: A communication channel is established between the main chain (100) and the shard chain (200) through the IBC protocol to calculate the Merkle tree root hash value (Hash_V) of the shard chain data and anchor it to the main chain to generate a cross-chain verification certificate (Cross-ChainCert) containing a timestamp.

[0028] Federated learning model construction: Secure multi-party computing (MPC) is used to encrypt and aggregate the local gradients of the shard chain. The dynamic credit assessment model (300) includes a basic rule layer (310) and a dynamic parameter layer (320). The driver chain (220) sends the credit repair record hash value (0x3a7b...e9f2) of a driver (ID number: XXX) to the main chain. After verification, the main chain generates a certificate number CC-20230820-0023 and uses it as a trusted data source for federated learning model training.

[0029] S3. Verify the cross-chain verification certificate through the federated learning module. If the verification is passed, use secure multi-party computing technology to encrypt and aggregate the local training gradients of each shard chain based on the cross-chain data to generate a unified dynamic credit assessment model.

[0030] Optionally, a smart contract trigger mechanism: a traffic violation handling contract in the smart contract (400) deployed in the vehicle chain sets a preset trigger condition (such as speeding value ≥ 20%). When the traffic behavior data meets the threshold, the dynamic credit assessment model is automatically called.

[0031] Encryption calculation process: ZK-SNARKs is used to verify the authenticity of the data source, and the input data is desensitized through Paillier homomorphic encryption. Output format: {score: 85, credibility index: 0.92, timestamp: 202308201503}.

[0032] For example, historical violation data (encrypted format: Encrypted_Violation_Data) is extracted from the vehicle chain, and a dynamic credit assessment model is called to calculate the driver's credit score (output ciphertext: 0x891a...c3d7), and the score result is written into a new block (Block#7821) of the driver chain.

[0033] S4. Control the real-time smart contract on the shard chain to call the dynamic credit assessment model to perform credit score calculation while capturing the latest traffic behavior data, and write the score result into a new block of the corresponding shard chain.

[0034] Optional feedback adjustment mechanism: The shard chain monitors the credit score deviation rate in real time (Δ=|actual value - predicted value| / maximum value). When Δ>5%, it sends a parameter adjustment request to the main chain (including the deviation data sample). The main chain updates the model weight matrix through the Q-learning algorithm. Cross-chain parameter synchronization: Update package structure: {parameter ID: W_0021, new weight value: 0.18, effective time: 202308210000}, using a double-signature verification mechanism to ensure transmission security.

[0035] For example, when the deviation rate of pedestrian violation scores for failure to yield to pedestrians in a certain area (such as Chaoyang District, Beijing) exceeds 7% for three consecutive days: the vehicle chain generates an adjustment request: {area code: 110105, deviation data: 0x5b3d...a9f1}; the main chain calculates the new score weight (original weight 0.15 → new weight 0.21); and the updated parameters are synchronized to the relevant shard chains through a cross-chain credit data storage request (TxID: XC-202308201605).

[0036] In an exemplary embodiment, the credit data of traffic participants are classified and stored in the corresponding shard chains in the multi-chain collaborative architecture, including: writing vehicle violation data into the vehicle chain, writing driver credit data into the driver chain, and writing transportation enterprise data into the enterprise chain.

[0037] It is understandable that a multi-chain collaborative architecture is constructed, including a main chain (100) and three shard chains: Vehicle chain (210): dedicated to storing vehicle dynamic behavior data, including but not limited to: violation type (running a red light, speeding, etc.); violation timestamp (accurate to milliseconds); geographic location coordinates (WGS84 standard).

[0038] Driver chain (220): stores driver credit files, including: driver's license status (normal / temporarily withheld / revoked); credit repair operation records (repair time, repair type); historical score change log.

[0039] Enterprise Chain (230): records the operational data of transport enterprises, including: complaint handling records (complaint type, response time); safety accident history (accident level, compensation ratio); service quality rating (monthly / quarterly rating).

[0040] Furthermore, the credit data of traffic participants are classified and stored in the corresponding shard chains in the multi-chain collaborative architecture, including: verifying the credit data storage request, and packaging the encrypted data into the block of the corresponding shard chain after meeting the preset credit data integrity verification conditions.

[0041] Among them, the encrypted data includes the timestamp data of the frequency of running red lights in the vehicle chain, the credit repair operation records in the driver chain, and the complaint response time log in the enterprise chain.

[0042] It's understandable that vehicle chain data includes timestamp data on a vehicle's red light running frequency. This data records the specific time each time a vehicle runs a red light, which is important for assessing the frequency of a vehicle's traffic violations. For example, if a vehicle runs red lights multiple times at different times within a month, these timestamp data can clearly reflect the temporal distribution of its violations.

[0043] Driver Chain Data: Contains a driver's credit repair records. When a driver's credit is damaged due to a traffic violation, they may repair their credit by participating in traffic safety education courses, volunteering, and other activities. Records of these actions will be stored in the Driver Chain. For example, if driver Zhang San participated in a traffic safety training course after violating the law, information such as the registration time, training time, and assessment results will be stored in the Driver Chain as a credit repair operation record.

[0044] Enterprise Chain Data: Complaint response time logs for companies. In the transportation sector, companies may receive complaints from users, such as taxi companies receiving complaints from passengers about their drivers' service attitudes. The company's response time to these complaints is recorded, forming a complaint response time log. For example, if a taxi company responds and handles a passenger complaint within 24 hours, the 24-hour response time will be recorded in the Enterprise Chain.

[0045] When new credit data of a traffic participant is generated and a credit data storage request is initiated, the system will first verify the credit data storage request. The verification process mainly includes the following aspects: Data source verification: Confirm whether the data comes from a legitimate data source, such as whether the vehicle's red light running data is accurately collected by the traffic management department's monitoring equipment, and whether the driver's credit repair operation records are provided by the relevant training institution or management department.

[0046] Format Verification: Checks whether the data conforms to the pre-set format requirements. For example, red light running frequency timestamp data should use a unified time format (e.g., YYYY - MM - DD HH:MM:SS), and credit repair operation records should include necessary information fields (e.g., operation type, operation time, operation result, etc.).

[0047] Permission verification: Ensure that the user or organization initiating the credit data storage request has the appropriate permissions to store that type of data. For example, only the traffic management department is authorized to upload red light running data, and the driver himself or the organization authorized by him can upload credit repair operation records.

[0048] After completing the credit data storage request verification, the system will perform a completeness check on the credit data. The completeness check conditions include but are not limited to: Data consistency check: Check the logical consistency within the data, such as whether the time sequence in the red light running frequency timestamp data is reasonable, and whether the operation results in the credit repair operation records match the operation type.

[0049] Hash value verification: The data is hashed and compared with the pre-stored hash value. If the hash values ​​are consistent, it means that the data has not been tampered with during transmission and storage.

[0050] Once the credit data storage request is verified and meets the pre-set credit data integrity verification conditions, the system will encrypt the data. The encryption algorithm can use common symmetric encryption algorithms (such as AES) or asymmetric encryption algorithms (such as RSA) to ensure data security.

[0051] The encrypted data will be packaged into the blocks of the corresponding shard chain. The specific process is as follows: Data classification: According to the type of data, the red light running frequency timestamp data of the vehicle chain is packaged into the block of the vehicle chain, the credit repair operation record of the driver chain is packaged into the block of the driver chain, and the complaint response time log of the enterprise chain is packaged into the block of the enterprise chain.

[0052] Block Generation: Each shard chain generates a new block based on its own consensus mechanism and appends the encrypted data to the block. For example, the vehicle chain uses a Proof-of-Work (PoW) consensus mechanism. When miners complete a certain computational task, a new block is generated and the timestamp of the red light running frequency is appended to the block.

[0053] Blockchain Linking: Newly generated blocks are linked to the blockchain of the corresponding shard chain, forming an immutable distributed ledger. In this way, the credit data of transportation participants is securely and accurately classified and stored in the corresponding shard chain in the multi-chain collaborative architecture.

[0054] In this embodiment, the classified storage method improves data management efficiency and query speed, while ensuring data accuracy and security through credit data storage request verification and data integrity verification. In addition, the encrypted data storage method further protects the privacy and data security of traffic participants.

[0055] In an exemplary embodiment, based on cross-chain data, secure multi-party computing technology is used to encrypt and aggregate the local training gradients of each shard chain to generate a unified dynamic credit assessment model, including: The encrypted red light running frequency timestamp data is extracted from the vehicle chain to generate a violation feature vector including a time decay factor.

[0056] Extract the encrypted red light running frequency timestamp data from the vehicle chain. Since the data in the vehicle chain is stored encrypted, it is necessary to use the corresponding decryption key for decryption.

[0057] Considering the timeliness of traffic violations, a time decay factor is introduced. For example, the closer the red light violation is to the current time, the greater the impact on the credit assessment, while the longer the red light violation is, the less impact it will have. The time decay factor can be defined as α ( t )= e − λt ,in t is the difference between the time when the red light running behavior occurs and the current time, λ is the attenuation coefficient.

[0058] According to the extracted timestamp data and time decay factor, generate the violation feature vector X containing the time decay factor v For example, if at different time points t 1, t 2,⋯, tn When a red light running behavior occurs, the corresponding time attenuation factors are α ( t 1), α ( t 2),⋯, α ( tn ), then the violation feature vector can be expressed as X v =[ α ( t 1), α ( t 2),⋯, α ( tn )].

[0059] The credit repair operation records are parsed from the driver chain to generate a behavioral feature vector containing the repair time weight.

[0060] Credit repair operation records are parsed from the driver chain. These records contain various actions taken by the driver to repair their credit, such as participating in traffic safety training, volunteering, etc. Each credit repair operation is assigned a repair timeliness weight. For example, a credit repair operation completed quickly may have a higher weight because it indicates that the driver has a strong willingness and ability to repair their credit. The repair timeliness weight can be determined based on factors such as the time the operation was completed and the type of operation. β Based on the parsed credit repair operation records and repair time weights, a behavioral feature vector X containing the repair time weights is generated. d For example, if the driver performs m The corresponding credit repair time weights for each credit repair operation are β 1, β 2,⋯, βm , then the behavior feature vector can be expressed as X d =[ β 1,β 2,⋯, βm ].

[0061] Complaint response time logs are analyzed from the enterprise chain to generate a service quality feature vector based on response speed.

[0062] Analyze the complaint response time log from the enterprise chain. This log records the enterprise's response time to user complaints and reflects the enterprise's service quality. Generate a service quality feature vector based on the response speed. For example, the response time can be quantified, such as dividing the response time into different levels, each level corresponding to a score, and then determine its value in the service quality feature vector based on the enterprise's actual response time. Suppose the enterprise's complaint response time is T , according to the preset duration level classification, determine the corresponding score s ( T ), then the service quality feature vector X e It can be expressed as X e =[ s ( T )].

[0063] The violation feature vector, behavior feature vector, and service quality feature vector are input into the federated learning model, fused through the spatiotemporal weight matrix, and the dynamic credit evaluation function is output.

[0064] The violation feature vector X v , behavioral feature vector X d and the service quality feature vector X e Input the federated learning model. This model can be a machine learning model such as a deep neural network or support vector machine. These feature vectors are fused using the spatiotemporal weight matrix W. The spatiotemporal weight matrix W takes into account the importance of different features in time and space. For example, in certain time periods or specific areas, violations may have a greater impact on credit assessments, and the corresponding weights will be higher. After training and fusion of the federated learning model, the dynamic credit assessment function is output. f (X v ,X d ,X e ;W).

[0065] Based on the output dynamic credit assessment function, a unified dynamic credit assessment model is constructed.

[0066] Dynamic credit assessment function based on output f (X v ,X d ,X e ;W), building a unified dynamic credit assessment model MThe model can input corresponding feature vectors according to different traffic participants and output their corresponding credit assessment results.

[0067] Optionally, the real-time smart contract on the control shard chain calls the dynamic credit assessment model to perform credit score calculations while capturing the latest traffic behavior data, including: When the real-time smart contract on the control shard chain receives a traffic behavior event trigger, it extracts the time decay factor and repair time weight in the dynamic credit evaluation function; (42) Combined with the spatiotemporal attributes of the current event, the encrypted scoring parameters are generated through homomorphic encryption calculation; (43) The encrypted scoring parameters are matrix-operated with the dynamic credit evaluation function to output the credit score ciphertext containing a verifiable timestamp.

[0068] It is understandable that (41) is used to extract key parameters to control the real-time smart contract on the shard chain to start the credit score calculation process when it receives a traffic behavior event trigger, such as a new red light run by a vehicle, a new credit repair operation by a driver, or a new complaint received by the enterprise. The time decay factor and repair time weight are extracted from the dynamic credit evaluation function. These parameters will be used in the subsequent credit score calculation.

[0069] This algorithm is used to generate encrypted scoring parameters. This algorithm combines the temporal and spatial attributes of the current event, such as the time and location of the event, and generates these encrypted scoring parameters through homomorphic encryption. Homomorphic encryption allows specific mathematical operations to be performed on ciphertext without decrypting the data, thus ensuring data privacy. For example, let the temporal and spatial attributes of the current event be S. Using the homomorphic encryption algorithm E, the relevant parameters are encrypted to generate the encrypted scoring parameter C = E(S).

[0070] Outputting the credit score ciphertext: Perform matrix operations on the encrypted score parameter C and the dynamic credit assessment function. Since the dynamic credit assessment function is constructed based on the previously generated eigenvectors and weight matrix, the credit score ciphertext can be obtained through matrix operations. Outputting the credit score ciphertext with a verifiable timestamp. The verifiable timestamp is used to prove the calculation time of the credit score, ensuring the timeliness and credibility of the score results. For example, the credit score ciphertext can be expressed as Y=[ Cscore , timestamp ],in Cscore is the encrypted credit score, timestampThis is a verifiable timestamp. This embodiment utilizes secure multi-party computation technology to encrypt and aggregate the local training gradients of each shard chain, leveraging cross-chain data to construct a unified dynamic credit assessment model that more comprehensively and accurately reflects the credit status of traffic participants. Furthermore, a real-time smart contract utilizes this model to calculate credit scores when capturing the latest traffic behavior data, ensuring the dynamic and timely nature of credit assessments.

[0071] In an exemplary embodiment, after writing the scoring result into a new block of the corresponding shard chain, the method further includes: based on the credit scoring deviation data monitored in real time by the shard chain, triggering the main chain to adjust the local parameters of the dynamic credit assessment model through the parameter optimization algorithm of the federated learning model, and synchronizing the updated parameters to the relevant shard chain through the cross-chain protocol.

[0072] Optionally, after writing the scoring results to a new block on the corresponding shard chain, the shard chain will monitor credit score deviation data in real time. For example, the vehicle chain can determine whether there is scoring deviation by comparing the vehicle's actual traffic behavior (such as the frequency and severity of recent violations) with the credit status reflected by the current credit score. If a vehicle with a previously high credit score is found to frequently run red lights within a short period of time, but the current credit score does not promptly reflect this change, then a credit score deviation has occurred.

[0073] When a credit score deviation is detected, the shard chain triggers the main chain to process the data. The main chain uses the parameter optimization algorithm of the federated learning model to adjust the local parameters of the dynamic credit assessment model. The federated learning model's parameter optimization algorithm will try different parameter adjustment strategies based on the credit score deviation to find the optimal parameter adjustment solution, allowing the model to more accurately reflect the credit status of traffic participants. For example, to address the scoring deviation caused by frequent red light running in the vehicle chain, the federated learning model's parameter optimization algorithm may increase the weight parameter of red light running in the credit assessment, making the model more sensitive to this type of violation.

[0074] After the main chain obtains the adjusted local parameters of the dynamic credit assessment model through the parameter optimization algorithm of the federated learning model, it synchronizes these updated parameters to the relevant shard chains via the cross-chain protocol. The cross-chain protocol ensures the secure and accurate transmission of data between different blockchains. For example, in the vehicle chain, the main chain sends the updated parameters related to vehicle violations to the vehicle chain. Upon receiving the updated parameters, the vehicle chain stores them and prepares to apply them to optimize its own credit assessment.

[0075] Furthermore, after synchronizing the updated parameters to the relevant shard chains through the cross-chain protocol, the method also includes: controlling the shard chains to perform back-calculation on the historical encrypted scoring parameters based on the updated dynamic credit assessment function; according to the back-calculation results, activating the correction parameters and updating the spatiotemporal dimension weight matrix of the federated learning model.

[0076] Optionally, based on the backtracking calculation results, the correction parameters are activated and the spatiotemporal dimension weight matrix of the federated learning model is updated, including: differentially comparing the backtracking calculation results with the expected correction value of the main chain to generate a weight matrix correction verification report; when the differential value is less than the preset tolerance, the correction parameters are activated and the spatiotemporal dimension weight matrix of the federated learning model is updated.

[0077] After the updated parameters are synchronized to the relevant shard chains via the cross-chain protocol, the shard chains retroactively calculate the historical encrypted scoring parameters based on the updated dynamic credit evaluation function. For example, the vehicle chain uses the updated dynamic credit evaluation function to recalculate the credit score corresponding to the previously stored encrypted scoring parameters on-chain. Suppose that during a certain period of time, the credit score of vehicle A was calculated based on the old model parameters. Now, the vehicle chain uses the updated dynamic credit evaluation function and combines all traffic behavior data of vehicle A during that period (such as red light running timestamp data) to recalculate its credit score.

[0078] The back-calculated results are differentially compared with the expected corrections from the main chain to generate a weight matrix correction verification report. Based on the overall objectives of traffic credit assessment and historical experience, the main chain predetermines expected corrections to credit scores under specific circumstances. For example, the main chain anticipates that after adjusting the weighting parameters for red light running, Vehicle A's credit score will decrease by a certain amount within a certain time period. The vehicle chain compares Vehicle A's new credit score, obtained through back-calculation, with the expected corrections from the main chain and calculates the difference. If the difference between the reduction in Vehicle A's back-calculated credit score and the expected correction from the main chain is x, the vehicle chain will be considered.

[0079] When the difference value falls below the preset tolerance, the correction parameter is activated and the federated learning model's spatiotemporal weight matrix is ​​updated. The preset tolerance is a threshold set based on the actual application scenario and the required accuracy of the credit assessment. Assuming the preset tolerance is \(\epsilon\), if \(x < \epsilon\), the difference between the back-calculated result and the expected correction value on the main chain is considered to be within an acceptable range, and the correction parameter is activated. The correction parameter updates the spatiotemporal weight matrix of the federated learning model based on the back-calculated result and the expected correction value on the main chain. For example, in the spatiotemporal weight matrix, the weights for red light running at different times and locations are adjusted, allowing the model to more accurately reflect the credit status of traffic participants in future credit assessments.

[0080] In this embodiment, by back-calculating historical data and updating the weight matrix, the federated learning model is further optimized, thereby improving the performance and reliability of the entire traffic credit assessment system.

[0081] In an exemplary embodiment, after activating the correction parameters and updating the spatiotemporal weight matrix of the federated learning model based on the backtracking calculation results, the method further includes: (a) When a credit repair operation record is detected on the driver chain, the main chain is triggered to update the repair time weight of the dynamic credit evaluation function.

[0082] Optionally, at a certain point, the federated learning model has completed updating the spatiotemporal weight matrix, and the dynamic credit assessment function generated based on this is being used to calculate the credit scores of traffic participants. For example, driver Li Ming's credit score is based on the current dynamic credit assessment function, which takes into account his driving violation record, credit repair efforts (if any), and the relevant data of his associated vehicles and companies. At this point, Li Ming's credit score is 70 (out of 100), his associated vehicle's credit score is 75, and his associated company's credit score is 80.

[0083] (b) Send weight update instructions to the vehicle chain and enterprise chain through the cross-chain protocol to synchronously adjust the credit scoring parameters of the associated entities.

[0084] Some time later, the Driver Chain detected that Li Ming had performed a credit repair operation. He had participated in traffic safety training and obtained a certificate. This credit repair operation was recorded in real time. After detecting Li Ming's credit repair operation, the Driver Chain sent a trigger signal to the main chain. Upon receiving the signal, the main chain initiated the process of updating the repair timeliness weight of the dynamic credit assessment function. Based on pre-set rules and algorithms, the main chain adjusted the repair timeliness weight based on factors such as the type of Li Ming's credit repair operation (traffic safety training) and completion time. For example, the original credit repair timeliness weight was 0.3. Considering the importance of traffic safety training in improving driver safety awareness and Li Ming's quick completion of the training and receipt of the certificate, the main chain increased the repair timeliness weight to 0.4.

[0085] After the main chain updates the repair timeliness weight, it sends a weight update instruction to the vehicle chain and enterprise chain via the cross-chain protocol. Upon receiving the instruction, the vehicle chain and enterprise chain simultaneously adjust the credit score parameters associated with Li Ming. For example, the vehicle chain recalculates the credit score parameters of vehicles associated with Li Ming based on the new repair timeliness weight. Because Li Ming's credit repair operation may have a positive impact on the vehicle's credit assessment, the vehicle chain adjusts the vehicle credit score calculation parameters associated with Li Ming, such as increasing the weight associated with Li Ming's credit status. The enterprise chain also makes similar adjustments, reassessing the credit association parameters between Li Ming and related enterprises.

[0086] (c) Input the updated scoring parameters into the federated learning model to generate a new version of the dynamic credit assessment function.

[0087] After the vehicle and enterprise chains adjusted their credit scoring parameters, they fed the updated scoring parameters into the federated learning model. The federated learning model, incorporating these updated parameters, recalculates and retrains, generating a new version of the dynamic credit assessment function. For example, the new dynamic credit assessment function prioritizes the impact of credit repair when calculating a driver's credit score and more accurately reflects changes in the credit scores of linked vehicles and businesses. Based on the new dynamic credit assessment function, Li Ming's credit score increased to 75, the linked vehicle's credit score increased to 80, and the linked business's credit score increased to 82.

[0088] In an exemplary embodiment, Figure 2 As shown, a decentralized transportation credit management device based on blockchain is provided, including: Data acquisition module 11 is used to classify and store the credit data of traffic participants in the corresponding shard chains in the multi-chain collaborative architecture; the shard chains include vehicle chains, driver chains, and enterprise chains; The multi-center verification module 12 is used to perform two-way anchor verification on the credit data hash value between the main chain and the shard chain, generate a cross-chain verification certificate, and synchronize it to the federated learning module of the main chain; The unified model training module 13 is used to verify the cross-chain verification certificate through the federated learning module. If the verification is passed, the local training gradients of each shard chain are encrypted and aggregated using secure multi-party computing technology based on the cross-chain data to generate a unified dynamic credit assessment model. The credit scoring module 14 is used to control the real-time smart contract on the shard chain to call the dynamic credit assessment model to perform credit scoring calculations while capturing the latest traffic behavior data, and write the scoring results into a new block of the corresponding shard chain.

[0089] Each module in the aforementioned blockchain-based decentralized transportation credit management device can be implemented in whole or in part through software, hardware, or a combination thereof. Each module can be embedded in or independent of a server processor in hardware form, or stored in a server memory in software form, allowing the processor to call and execute the corresponding operations of each module.

[0090] In an exemplary embodiment, a server is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented: The credit data of traffic participants are classified and stored in the corresponding shard chains in the multi-chain collaborative architecture; the shard chains include vehicle chains, driver chains and enterprise chains; Perform two-way anchor verification on the credit data hash value between the main chain and the shard chain, generate a cross-chain verification certificate and synchronize it to the federated learning module of the main chain; The cross-chain verification certificate is verified through the federated learning module. If the verification passes, the local training gradients of each shard chain are encrypted and aggregated using secure multi-party computing technology based on the cross-chain data to generate a unified dynamic credit assessment model. The real-time smart contract on the control shard chain captures the latest traffic behavior data, calls the dynamic credit assessment model to perform credit score calculation, and writes the score result into a new block of the corresponding shard chain.

[0091] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented: The credit data of traffic participants are classified and stored in the corresponding shard chains in the multi-chain collaborative architecture; the shard chains include vehicle chains, driver chains and enterprise chains; Perform two-way anchor verification on the credit data hash value between the main chain and the shard chain, generate a cross-chain verification certificate and synchronize it to the federated learning module of the main chain; The cross-chain verification certificate is verified through the federated learning module. If the verification passes, the local training gradients of each shard chain are encrypted and aggregated using secure multi-party computing technology based on the cross-chain data to generate a unified dynamic credit assessment model. The real-time smart contract on the control shard chain captures the latest traffic behavior data, calls the dynamic credit assessment model to perform credit score calculation, and writes the score result into a new block of the corresponding shard chain.

[0092] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the following steps: The credit data of traffic participants are classified and stored in the corresponding shard chains in the multi-chain collaborative architecture; the shard chains include vehicle chains, driver chains and enterprise chains; Perform two-way anchor verification on the credit data hash value between the main chain and the shard chain, generate a cross-chain verification certificate and synchronize it to the federated learning module of the main chain; The cross-chain verification certificate is verified through the federated learning module. If the verification passes, the local training gradients of each shard chain are encrypted and aggregated using secure multi-party computing technology based on the cross-chain data to generate a unified dynamic credit assessment model. The real-time smart contract on the control shard chain captures the latest traffic behavior data, calls the dynamic credit assessment model to perform credit score calculation, and writes the score result into a new block of the corresponding shard chain.

[0093] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0094] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the 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-mentioned methods. Among them, any reference to memory, database or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the various embodiments provided herein may be, but are not limited to, general-purpose processors, central processing units (CPUs), graphics processing units (GPUs), digital signal processors (DSPs), programmable logic devices (PLDs), quantum computing-based data processing logic devices, artificial intelligence (AI) processors, and the like.

[0095] The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0096] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A decentralized transportation credit management method based on blockchain, characterized in that: include: The credit data of traffic participants are classified and stored in the corresponding shard chains in the multi-chain collaborative architecture; the shard chains include vehicle chains, driver chains and enterprise chains; Perform two-way anchor verification on the credit data hash value between the main chain and the shard chain, generate a cross-chain verification certificate and synchronize it to the federated learning module of the main chain; The cross-chain verification certificate is verified by the federated learning module. If the verification passes, the local training gradients of each shard chain are encrypted and aggregated using secure multi-party computing technology based on the cross-chain data to generate a unified dynamic credit assessment model. The real-time smart contract on the shard chain is controlled to call the dynamic credit assessment model to perform credit score calculation while capturing the latest traffic behavior data, and write the score result into a new block of the corresponding shard chain.

2. The method according to claim 1, characterized in that The process of classifying and storing the credit data of traffic participants in corresponding shard chains in the multi-chain collaborative architecture includes: Write vehicle violation data into the vehicle chain, write driver credit data into the driver chain, and write transportation enterprise data into the enterprise chain.

3. The method according to claim 2, characterized in that The process of classifying and storing the credit data of traffic participants in corresponding shard chains in the multi-chain collaborative architecture includes: Verify the obtained credit data storage request, and after meeting the preset credit data integrity verification conditions, package the encrypted data into the block of the corresponding shard chain; Among them, the encrypted data includes the red light running frequency timestamp data of the vehicle chain, the credit repair operation record of the driver chain, and the complaint response time log of the enterprise chain.

4. The method according to claim 2, characterized in that Based on cross-chain data, secure multi-party computing technology is used to encrypt and aggregate the local training gradients of each shard chain to generate a unified dynamic credit assessment model, including: Extracting encrypted red light running frequency timestamp data from the vehicle chain to generate a violation feature vector including a time decay factor; Parsing credit repair operation records from the driver chain to generate a behavior feature vector including a repair timeliness weight; Analyze complaint response time logs from the enterprise chain to generate a service quality feature vector based on response speed; Input the violation feature vector, behavior feature vector, and service quality feature vector into the federated learning model, fuse them through the spatiotemporal weight matrix, and output a dynamic credit evaluation function; Based on the output dynamic credit assessment function, a unified dynamic credit assessment model is constructed.

5. The method according to claim 4, characterized in that The real-time smart contract on the shard chain is controlled to call the dynamic credit assessment model to perform credit score calculation while capturing the latest traffic behavior data, including: Controlling the real-time smart contract on the shard chain to extract the time decay factor and repair time weight in the dynamic credit evaluation function when receiving a traffic behavior event trigger; Combined with the spatiotemporal attributes of the current event, encrypted scoring parameters are generated through homomorphic encryption calculations; Perform matrix operations on the encrypted scoring parameters and the dynamic credit assessment function to output a credit score ciphertext containing a verifiable timestamp.

6. The method according to claim 4, characterized in that After writing the scoring result into a new block of the corresponding shard chain, the method further includes: Based on the credit score deviation data monitored in real time by the shard chain, the main chain is triggered to adjust the local parameters of the dynamic credit assessment model through the parameter optimization algorithm of the federated learning model, and the updated parameters are synchronized to the relevant shard chain through the cross-chain protocol.

7. The method according to claim 6, characterized in that After synchronizing the updated parameters to the relevant shard chains through the cross-chain protocol, the method further includes: Controlling the shard chain to perform retrospective calculations on historical encrypted scoring parameters based on the updated dynamic credit assessment function; According to the backtracking calculation results, the correction parameters are activated and the spatiotemporal dimension weight matrix of the federated learning model is updated.

8. The method according to claim 7, characterized in that The step of activating the correction parameters and updating the spatiotemporal weight moments of the federated learning model according to the backtracking calculation results includes: Perform differential comparison between the backtracking calculation results and the expected correction value of the main chain to generate a weight matrix correction verification report; When the difference value is less than a preset tolerance, the correction parameter is activated and the spatiotemporal dimension weight matrix of the federated learning model is updated.

9. The method according to claim 4, characterized in that After activating the correction parameters and updating the spatiotemporal dimension weight matrix of the federated learning model according to the backtracking calculation results, the method further includes: When the driver chain credit repair operation record is monitored, the main chain is triggered to update the repair time weight of the dynamic credit evaluation function; Send weight update instructions to the vehicle chain and enterprise chain through the cross-chain protocol to synchronously adjust the credit score parameters of the associated entities; The updated scoring parameters are input into the federated learning model to generate a new version of the dynamic credit assessment function.

10. A decentralized transportation credit management device based on blockchain, characterized in that: include: A data acquisition module is used to classify and store the credit data of traffic participants in corresponding shard chains in the multi-chain collaborative architecture; the shard chains include vehicle chains, driver chains, and enterprise chains; The multi-center verification module is used to perform two-way anchor verification on the credit data hash value between the main chain and the shard chain, generate cross-chain verification certificates and synchronize them to the federated learning module of the main chain; A unified model training module is used to verify the cross-chain verification certificate through the federated learning module. If the verification is successful, the module uses secure multi-party computing technology to encrypt and aggregate the local training gradients of each shard chain based on the cross-chain data to generate a unified dynamic credit assessment model. The credit scoring module is used to control the real-time smart contract on the shard chain to call the dynamic credit assessment model to perform credit scoring calculations while capturing the latest traffic behavior data, and write the scoring results into a new block of the corresponding shard chain.

Images