A compliant tax refund method and intelligent declaration system for cross-border e-commerce based on blockchain-based evidence storage

Abstract

This invention relates to the field of computer data processing technology, and discloses a method and intelligent declaration system for compliant tax refunds in cross-border e-commerce based on blockchain evidence storage. The method constructs a benchmark library containing the physical attributes and fair density ranges of goods; acquires order transaction, logistics measurement, and fund time-series data; performs multi-dimensional cross-validation based on the benchmark library, including physical spatial logic verification based on geometric packing simulation, attribute consistency verification based on density distribution matching, and joint quality and time-series verification; determines the risk level based on the comprehensive results; generates digital fingerprints for compliant data and writes them into a distributed ledger; and generates declaration documents using on-chain indexes. This invention verifies the authenticity of trade through physical simulation and multi-dimensional logic, and combines blockchain technology to ensure the immutability of evidence, effectively improving the compliance and review efficiency of tax refund declarations.

Description

Technical Field

[0001] This invention relates to the field of computer data processing technology, specifically to a compliant tax refund method and intelligent declaration system for cross-border e-commerce based on blockchain-based evidence storage. Background Technology

[0002] Cross-border e-commerce export trade has become a significant driver of foreign trade growth. Export tax rebates, as a key measure supporting this industry, allow companies to receive refunds of value-added tax and consumption tax already paid in the production or distribution stages. The core of export tax rebate management lies in verifying the authenticity of the trade background to prevent fraudulent tax evasion. Currently, tax authorities and related service platforms primarily conduct compliance audits by comparing the consistency of electronic data among order information, logistics waybills, and payment records. This "three-way comparison" model focuses on verifying whether textual information such as transaction time, amount, and entity name matches.

[0003] However, existing data verification technologies have significant limitations in deeply verifying the authenticity of trade logic. Traditional auditing models primarily focus on checking the consistency of surface information on documents, lacking effective means to verify the consistency between digital declaration information and the physical logistics entity. In practice, there are cases of tax evasion by fabricating high-value commodity transactions but actually shipping low-value goods or empty packages. The data generated by such activities often appears to match superficially at the document level, but has loopholes in physical attributes and economic logic. For example, the declared volume of goods may not match the actual volume of the logistics package, or the paid logistics fees may be disproportionate to the value of the goods, violating normal commercial economic rationality.

[0004] Especially for complex orders containing a mix of various goods, existing technologies struggle to accurately determine the authenticity of the package's contents based on a single weight or volume parameter. Lacking the ability to calculate physical models for mixed SKU (stock keeping unit) packing, the system cannot identify covert fraudulent activities such as "small items in large packages" or abnormal density. Furthermore, existing declaration data primarily relies on centralized server storage, which carries the risk of data tampering or replacement during generation, transmission, and review. This results in insufficient credibility and non-repudiation of the electronic evidence chain ultimately submitted to tax authorities, increasing the difficulty of review and risk management costs for regulatory departments. Summary of the Invention

[0005] This invention aims to solve the technical problems in the existing cross-border e-commerce export tax refund application process, such as the difficulty in verifying the authenticity of transaction data, the limited means of verifying logistics and capital flow logic, and the ease with which data can be tampered with.

[0006] To address the above problems, the present invention provides the following technical solution:

[0007] The first aspect of this invention provides a compliant tax refund method for cross-border e-commerce based on blockchain-based evidence storage.

[0008] This method first constructs a benchmark library containing commodity physical attribute parameters and fair density ranges of customs codes, providing a physical-level truth reference for subsequent verification. In response to cross-border transaction events, the system acquires order transaction data, actual logistics measurement data, and time-series financial data, achieving the aggregation of multi-source heterogeneous data.

[0009] Based on this, the method performs multidimensional cross-validation based on physical and economic laws, specifically including physical space logic verification, attribute consistency verification, and joint logic verification.

[0010] In the physical space logic verification, this method introduces geometric packing simulation technology, virtually loading the geometric shapes of goods in the order into the logistics package space to determine whether there are physical conflicts. If there are no conflicts, the theoretical minimum remaining void ratio of the package is further calculated, and the freight value ratio is calculated by combining the order transaction data and the logistics freight amount. By setting void ratio warning thresholds and freight ratio warning thresholds, this method can identify abnormal transactions that violate economic rationality, such as "small goods in large packages" or "low value with high freight," thereby effectively preventing tax fraud risks. To improve the accuracy of the simulation, the verification process can also introduce a deformation tolerance coefficient to compensate for the volume change of flexible packaging materials under pressure. This coefficient takes a value within a specific range based on the material type of the goods.

[0011] In attribute consistency verification, this method calculates the apparent density of logistics packages and compares it with the fair density range in a benchmark database. For complex orders containing multiple goods, the method calculates the theoretical volume ratio based on the declared quantity and standard physical dimensions of each good. This ratio is then used to weight and aggregate the fair density ranges for the corresponding customs codes of each good, generating a dynamic fair density range. This dynamic range generation mechanism solves the problem of difficult density verification for mixed packages, ensuring the accuracy of apparent density comparison.

[0012] In the joint logic verification, this method, based on the law of conservation of mass, adds the cumulative net weight of each item in the order to the weight of the packaging material estimated based on the actual measured dimensions to obtain the theoretical total weight, and verifies whether the deviation between this theoretical total weight and the actual measured total weight in logistics is within a preset tolerance range. Simultaneously, this method performs time-series logic and interval period verification on the time nodes of order generation, payment completion, and logistics pickup to ensure that the transaction process conforms to business practices.

[0013] After completing the above verifications, the method integrates the verification results. If any key indicators are abnormal, the transaction is directly rejected; otherwise, a weighted model is used to calculate a comprehensive compliance reputation value, and the risk assessment level of the transaction is determined accordingly. For compliant data that passes the risk assessment, the method performs standardized serialization processing, concatenates it with the reputation value and random salt value, and then performs a hash operation to generate a digital fingerprint. This digital fingerprint, along with the transaction serial number and timestamp, is written into the distributed ledger network, utilizing the immutability of blockchain to ensure the authenticity of the evidence data. Finally, by receiving the on-chain transaction hash returned by the smart contract as the evidence index, an export tax refund declaration detail table embedded with a digital signature or QR code is generated, achieving a strong correlation between the declaration data and the on-chain evidence.

[0014] The second aspect of this invention provides a smart declaration system for compliant tax refunds in cross-border e-commerce based on blockchain-based evidence storage.

[0015] The system includes a benchmark library management module, a data collection module, a verification module, a certificate storage module, and an application generation module.

[0016] The benchmark library management module is configured to build and maintain a benchmark library containing commodity physical attribute parameters and customs code fair density ranges. It provides basic data support for verification by establishing a physical vector mapping between SKUs and physical vectors containing length, width, height and quality parameters, and by determining fair density ranges based on historical data statistics.

[0017] The data collection module is configured to respond to cross-border transaction events by acquiring order transaction data, logistics measurement data, and fund time series data from different data sources to ensure the integrity of the verification data.

[0018] The verification module is the core processing component of the system, configured to perform multi-dimensional cross-validation on the collected data based on a benchmark library. This verification process encompasses physical spatial logic verification based on geometric binning simulation, attribute consistency verification based on density distribution matching, and joint logic verification based on mass conservation and time order. The verification module further integrates a multi-dimensional fusion calculation unit, which is responsible for synthesizing the results of various verifications, calculating a comprehensive compliance reputation value using a weighted model, and determining the risk assessment level of the transaction based on the threshold range to which the reputation value belongs.

[0019] The evidence storage module is configured to receive the verification results and key parameters output by the verification module, generate a digital fingerprint containing verification data, and call a smart contract to write the data into the distributed ledger network, ensuring the integrity and immutability of the evidence chain.

[0020] The declaration generation module is configured to respond to transaction data that has passed risk assessment by calling the evidence storage module to obtain the on-chain evidence storage index (such as the on-chain transaction hash), and bind the index with the original order information to generate an export tax refund declaration document with the evidence storage index and digital signature, thereby automating and ensuring compliance of the declaration process.

[0021] This invention provides a compliant tax refund method and intelligent declaration system for cross-border e-commerce based on blockchain-based evidence storage. It has the following beneficial effects:

[0022] 1. This invention achieves in-depth verification of the physical authenticity of cross-border transactions by constructing a benchmark library containing the physical attributes of goods and performing geometric packing simulation. Unlike traditional auditing methods that rely solely on comparing surface information on documents, this method calculates the theoretical minimum remaining void ratio of the package and combines this with the ratio of freight cost to the value of the goods for economic logic judgment. This dual verification mechanism based on physical space and economic rationality can effectively identify abnormal transactions where the declared goods do not match the actual volume of the logistics package, or where the freight cost ratio significantly violates conventional business logic, thereby discovering and blocking fraudulent trade or the use of low-value goods to fraudulently obtain tax refunds at the source.

[0023] 2. This invention proposes a dynamic fair density interval generation method based on theoretical volume ratio for complex orders containing multiple goods. This method calculates the theoretical volume weight of each good in the package based on the declared quantity and standard physical dimensions of each good in the order, and then weights and aggregates the fair density intervals corresponding to different customs codes accordingly. This approach solves the technical challenge of determining a unified density standard in mixed packaging scenarios, enabling the system to accurately determine whether the apparent density of the logistics package is within a reasonable composite density interval, thus improving the accuracy of attribute consistency verification for multi-SKU mixed packages.

[0024] 3. This invention tightly integrates the results of multi-dimensional cross-validation with blockchain evidence storage technology to construct an immutable export tax refund declaration evidence chain. The system serializes and hashes the verified transaction data, compliance reputation value, and random salt value to generate a digital fingerprint, which is then written into the distributed ledger network. The final generated export tax refund declaration details table directly embeds the transaction hash returned on the blockchain as an index, enabling the tax audit end to directly trace back to the verified original data state through this index. This ensures the authenticity and integrity of the declaration data throughout the entire process of generation, transmission, and storage, reducing the risk of data forgery. Attached Figure Description

[0025] Figure 1 This is a schematic diagram of the structure of a blockchain-based intelligent declaration system for compliant tax refunds in cross-border e-commerce, according to an embodiment of the present invention.

[0026] Figure 2This is a flowchart illustrating a blockchain-based compliant tax refund method for cross-border e-commerce, according to one embodiment of the present invention.

[0027] Figure 3 This is a visual schematic diagram of the multi-dimensional verification logic in an embodiment of the present invention.

[0028] Among them, 110 is the data collection module; 120 is the benchmark library management module; 130 is the verification module; 131 is the packing simulation unit; 132 is the density mapping unit; 133 is the time series analysis unit; 134 is the multi-dimensional fusion calculation unit; 140 is the evidence storage module; and 150 is the application generation module. Detailed Implementation

[0029] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0030] See attached document Figure 1 This invention provides a smart declaration system for compliant tax refunds in cross-border e-commerce based on blockchain-based evidence storage. The system includes: a data collection module 110, a benchmark database management module 120, a verification module 130, an evidence storage module 140, and a declaration generation module 150. These modules interact with each other via a system bus or network communication interface.

[0031] The data collection module 110 is configured to establish communication connections with external data sources, including enterprise resource planning systems, e-commerce platform order interfaces, logistics service provider data interfaces, and payment gateway interfaces. The data collection module 110 receives and parses multi-source heterogeneous data, including order transaction data, logistics measurement data, and time-series financial data. Specifically, order transaction data includes a list of product SKUs, declared quantities, and declared HS codes; logistics measurement data includes package outer packaging dimensions, total package weight, and billing freight information collected by the logistics-end DWS device.

[0032] The benchmark library management module 120 is configured to store and manage the basic physical parameters required for compliance verification. Internally, the benchmark library management module 120 maintains an SKU physical attribute database and an HS code density mapping table. The SKU physical attribute database stores the standard length, width, and height dimensions and standard mass data corresponding to each smallest inventory unit. The HS code density mapping table stores the fair physical density range corresponding to each customs code, and the fair physical density range is determined based on the statistical distribution of historical compliance data.

[0033] The verification module 130 is the core processing unit of the blockchain-based cross-border e-commerce compliant tax refund intelligent declaration system. It is configured to perform cross-verification of the data acquired by the data collection module 110 based on physical and economic logic according to a preset algorithm. The verification module 130 includes a packing simulation unit 131, a density mapping unit 132, a time series analysis unit 133, and a multi-dimensional fusion calculation unit 134.

[0034] The packing simulation unit 131 is configured to perform geometric space logic operations, map the geometric dimensions of the goods in the order transaction data to the package space in the actual logistics measurement data, calculate the theoretical minimum remaining void ratio, and make economic logic judgments based on the ratio of freight amount to goods value.

[0035] The density mapping unit 132 is configured to calculate the apparent density of the package and compare the apparent density with the fair physical density range in the HS code density mapping table to verify the authenticity of the declared commodity attributes.

[0036] The timing analysis unit 133 is configured to perform quality conservation calculations and timestamp logic checks to verify whether the deviation between the theoretical total mass and the measured total mass is within the tolerance range, and whether the timing logic of order generation time, payment time and logistics pickup time conforms to the preset rules.

[0037] The multidimensional fusion computing unit 134 is configured to receive various verification results output by the packing simulation unit 131, the density mapping unit 132, and the time series analysis unit 133, calculate the comprehensive compliance reputation value using a preset weighted model, and generate the final risk assessment level based on the reputation value.

[0038] The evidence storage module 140 is configured as the underlying trust component of the system. It responds to the request of the declaration generation module 150, receives the verification results and key parameters output by the verification module 130, generates a digital digest containing the verification process data using a hash algorithm, and writes the digital digest into the distributed ledger network to form an immutable compliance certificate.

[0039] The declaration generation module 150 is configured as the system's output terminal. Based on the compliant transaction data determined by the verification module 130, it extracts the declaration elements required by the tax authorities. At the same time, it calls the evidence storage module 140 to complete the data upload to the blockchain, obtain the on-chain evidence storage index, and finally generate an export tax refund declaration detail table with the evidence storage index attached.

[0040] See attached document Figure 2 This invention provides a method for compliant tax refunds in cross-border e-commerce based on blockchain-based evidence storage. This method is applied to the aforementioned intelligent declaration system for compliant tax refunds in cross-border e-commerce based on blockchain-based evidence storage, and includes the following steps:

[0041] S100, the benchmark library management module 120 constructs a multi-dimensional physical fingerprint benchmark library, establishes a mapping relationship between product SKU and physical vector, the physical vector includes length, width, height and quality parameters, and at the same time establishes a mapping relationship between customs code and fair density range;

[0042] S200, the data collection module 110 responds to cross-border transaction events, acquires and preprocesses transaction-related data in real time, extracts order transaction data groups, logistics measurement data groups and fund time series data groups from different data sources, and converts the above data groups into a vector format that the system can recognize;

[0043] S300, the packing simulation unit 131 in the verification module 130 performs physical space logic verification, attempts to virtually pack all the geometric objects of the order into the logistics package space. If there is a physical conflict, it is judged as abnormal. If there is no conflict, it further calculates the theoretical minimum remaining void ratio of the package and judges whether the packaging method conforms to economic rationality logic in combination with the logistics freight ratio.

[0044] S400, the density mapping unit 132 in the verification module 130 performs attribute consistency verification. It calculates the apparent density of the package based on the actual measured weight and volume of the logistics, and determines whether the apparent density falls within the fair density range corresponding to the declared customs code, thereby verifying the consistency of the physical attributes between the declared category and the actual shipped category.

[0045] S500, the timing analysis unit 133 in the verification module 130 performs joint quality and timing verification, calculates the absolute deviation between the theoretical total quality of the order and the actual total quality of the logistics, verifies whether it is less than the preset tolerance, and verifies whether the order time, payment time and logistics pickup time sequence and interval period conform to the preset logic.

[0046] S600, the multi-dimensional fusion calculation unit 134 in the verification module 130 calculates the comprehensive compliance reputation value using a weighted model based on the verification results of each dimension output by S300, S400 and S500, and generates a risk assessment level based on the reputation value.

[0047] S700, the declaration generation module 150 responds to the transaction data that has passed the risk assessment by calling the evidence storage module 140 to extract key verification parameters, generate digital fingerprints and store them on the chain. Then, based on the returned on-chain evidence storage index, it generates and outputs the export tax refund declaration document.

[0048] The following provides a detailed description of each step of the method of the present invention.

[0049] See attached document Figure 2Step S100 is executed by the benchmark library management module 120, aiming to provide standardized data references for subsequent physical space simulation and attribute verification. The benchmark library management module 120 interacts with the enterprise ERP system and the customs public database through a standardized data interface, transforming unstructured commodity attributes into system-computable vector data. Specifically, this includes the following steps:

[0050] Step S101: Construct the SKU physical attribute vector mapping table.

[0051] This step aims to digitize the physical form of the goods to facilitate geometric packing simulation by computer. The baseline library management module 120 iterates through each smallest SKU in the enterprise's inventory list, extracts its physical parameters, and encapsulates them into a physical attribute vector. For any given product... Its physical property vector The definition is as follows:

[0052] ;

[0053] In the formula, , , Representing the products Length, width, and height in standard packaging condition, in centimeters; Indicates goods Net weight per piece, in kilograms; This indicates the packaging deformation tolerance coefficient of the product.

[0054] This parameter is used to compensate for the volume change of flexible packaging materials under pressure during subsequent packing simulations. Its value is determined based on the material type of the product. The system has a pre-set material-coefficient mapping table for rigid materials (such as metals and rigid plastics). Set to 1.0; for flexible or compressible materials (such as textiles and down products). The compression ratio is set between 1.0 and 1.2. For example, when the SKU attribute is marked as "down jacket", the system automatically matches... =1.15.

[0055] The benchmark library management module 120 will store the above vectors An index is established and associated with the unique identifier of the product SKU, and stored in the SKU physical attribute database. For the specific storage structure and indexing technology of the database, those skilled in the art can use relational databases or key-value databases, which are well-known technologies in the field and will not be elaborated upon here.

[0056] Step S102: Construct a reference density distribution model based on HS coding.

[0057] This step is based on the principle that the physical density property of substances has relative stability, and uses the statistical laws of big data to define the reasonable physical range of various commodities. The benchmark library management module 120 establishes a physical density distribution model for various commodity categories based on the first six or eight digits of the customs code (HS Code). This step determines the fair range based on the statistical distribution of historical compliant export data.

[0058] For any customs code category The system selects historical transaction data of all completed tax refunds and no inspection abnormalities within this category within a preset time period (e.g., the most recent 12 months) as the sample set. To ensure the accuracy of density calculations, the sample set... Prioritize filtering historical orders that contain only a single HS code product, or mixed orders where the volume percentage of that HS code product exceeds a preset threshold (e.g., 90%), to eliminate interference from mixed-type products on density calculation.

[0059] For the sample set Each sample Calculate its historical apparent density :

[0060] ;

[0061] In the formula, Historical Sample The actual total weight measured in logistics. , , Historical Sample The actual measured dimensions of the outer packaging for logistics.

[0062] The benchmark library management module 120 calculates the sample set. density mean and standard deviation This allows for the determination of the customs code category. fair physical density range :

[0063] ;

[0064] In the formula, This is the confidence interval coefficient, and its value depends on the system's tolerance for false alarm rates. Typically, the system is set to... It covers approximately 95% of normal samples; in scenarios with stricter risk control requirements, it can be set... This covers approximately 99.7% of the samples.

[0065] The benchmark library management module 120 will calculate the interval With customs code A mapping relationship is established and stored in the HS code density mapping table. When a certain customs code lacks sufficient historical sample data, the system allows manual input of an industry-standard density range as an initial value, which is then dynamically updated during subsequent data accumulation. and .

[0066] Step S103: Perform regular cleaning and version control of the benchmark data.

[0067] Due to potential adjustments in product specifications and changes in market packaging materials, benchmark data needs to be kept up-to-date. The benchmark database management module 120 is configured with scheduled tasks to perform full or incremental updates to the SKU physical attribute database and HS code density mapping table at preset intervals. This updates are necessary when product parameters in the ERP system change, or when historical sample sets... When new compliance data is introduced, the module automatically recalculates the above vectors. and interval The verification module 130 generates a new data version number. During verification, it always calls the benchmark library version that matches the transaction occurrence time to ensure the time validity of the verification logic.

[0068] See attached document Figure 2 Step S200 is executed by the data collection module 110, aiming to transform the original transaction vouchers scattered across different business systems into feature vectors that the verification module can directly read. This data covers three core dimensions of the transaction: commercial agreement, physical entity, and fund flow, collectively forming the basic information source for verifying the authenticity of the trade. The data collection module 110 maintains communication connections with the Enterprise Resource Planning (ERP) system, the Logistics Service Provider Data Interface (DWS system), and the payment gateway interface through a pre-built API adapter or ETL (Extract, Transform, Load) tool, specifically including the following steps:

[0069] Step S201: Extract and parse the original data from multiple sources.

[0070] The data collection module 110 responds to the generation or status change event of cross-border e-commerce orders by pulling data from the three data sources mentioned above in parallel.

[0071] Extract basic order information from the ERP system, including the unique order number, the list of included product SKUs, the declared quantity of each SKU, and the corresponding declared unit price;

[0072] Extract actual logistics measurement information from the logistics service provider interface, including the waybill number, the length, width and height dimensions of the package outer packaging obtained by the DWS device, the total weight of the package recorded by the electronic scale, and the actual logistics cost incurred;

[0073] Extract fund flow information from the payment gateway, including the payment transaction number, payment completion timestamp, and actual amount received.

[0074] For parsing the communication protocols (such as RESTful API, SOAP) and data exchange formats (such as JSON, XML) of data interfaces, those skilled in the art can use existing network communication libraries and serialization tools, which are well-known technologies in the field and will not be elaborated here.

[0075] Step S202: Perform data cleaning and unit normalization.

[0076] Since different data sources may use different units of measurement or time formats, the data collection module 110 has built-in unit conversion logic. This module uniformly converts all length units to centimeters (cm), weight units to kilograms (kg), and currency amounts to the base currency (e.g., RMB) based on the exchange rate on the transaction date. The "transaction date" refers to the payment completion timestamp. For the corresponding calendar day, the exchange rate is the RMB central parity rate published by the People's Bank of China on that calendar day. If it is a non-working day, the data of the previous working day is used. If it involves currencies not listed in the central bank's central parity rate list, the spot buying rate of that currency against RMB published by a designated commercial bank (such as the Bank of China) on that day is used.

[0077] All timestamps are uniformly converted to UTC standard time or system local time. In addition, the module pre-validates data with missing values ​​or incorrect formats. For critical fields (such as package weight, payment amount, and quantity of goods), if there are missing values ​​or the format cannot be parsed, the data is directly marked as invalid and an exception log is triggered, and it does not enter the subsequent vectorization process. For non-critical fields (such as some remarks), missing values ​​are allowed, but they will be filled in the vector with specific symbols (such as NULL).

[0078] Step S203: Construct a feature vector group.

[0079] The data collection module 110 reorganizes the cleaned data into three-dimensional feature vectors: order transaction vector group, logistics measurement vector, and fund time series vector, which describe the key features of a transaction from the perspectives of commodity composition, physical form, and fund flow, respectively.

[0080] For order data in a single transaction, construct an order transaction vector group. Since an order may contain multiple items, Represented as a set:

[0081] ;

[0082] In the formula, This indicates the number of different SKUs included in the order; For the first A unique identifier for each product, used to index the SKU physical attribute database constructed in step S100; For the first The declared quantity of each type of commodity; For the first The declared unit price of the commodity; For the first Customs code for the declared commodity.

[0083] Based on the actual logistics measurement data, construct the logistics measurement vector. :

[0084] ;

[0085] In the formula, , , These represent the actual measured length, width, and height of the logistics package, respectively. This indicates the actual measured total weight of the package; This indicates the actual amount of logistics fees paid for the package.

[0086] Construct a time-series vector for funds and time data. :

[0087] ;

[0088] In the formula, This indicates the time the order was generated on the e-commerce platform; This indicates the payment completion time recorded by the payment gateway; This indicates the pickup and scanning time recorded by the logistics service provider; This indicates the actual payment amount recorded by the payment gateway.

[0089] The data aggregation module 110 will generate vector groups , and Package the data and attach a unique transaction serial number, then transmit it to the verification module 130 for use in subsequent steps.

[0090] Through the above vectorization process, the system eliminates the heterogeneity of the original data, enabling subsequent physical simulations and logical verifications to be performed based on a unified data model.

[0091] See attached document Figure 2Step S300 is executed by the packing simulation unit 131 in the verification module 130. It aims to identify false trade or abnormal declaration behavior by calculating the feasibility of physical space and the economic rationality of packaging methods. The core logic of this step is to verify whether the declared goods can actually be packed into the logistics package by utilizing the physical exclusivity of "materials occupying space" on the one hand; and to verify whether the merchant has abnormal shipping behavior that violates the principle of cost minimization by utilizing the "rational economic man" assumption on the other hand.

[0092] The packing simulation unit 131 calls the SKU physical attribute database constructed in step S100 and the vector data generated in step S200, specifically including the following steps:

[0093] Step S301: Perform geometric space feasibility verification.

[0094] Packing simulation unit 131 reads order transaction vector group All product SKUs and their quantities are retrieved from the SKU physical attribute database, along with their corresponding standard sizes. The unit constructs a virtual three-dimensional container. Considering that actual logistics packaging materials (such as corrugated cardboard) have physical thickness, the system needs to first calculate the effective internal dimensions of the container. Let the packaging wall thickness correction factor be... (Typically, the value is taken from 0.5cm to 1.0cm), then the effective internal size of the virtual container is... The calculation is as follows:

[0095] ;

[0096] Subsequently, a 3D packing algorithm was used to attempt to virtually place all the goods in the order into the available internal space.

[0097] In this process, a geometric verification function is defined. Its output is a Boolean value. :

[0098] ;

[0099] In the formula, Indicating orders This is a set of dimensions and quantities of various commodities after deformation tolerance correction. Among them, the corrected commodity dimensions... (Similarly, calculate the width and height), here we introduce... It is designed to simulate the volume compression of soft goods (such as down jackets) under pressure.

[0100] The algorithm considers the six degrees of freedom of rotation of the goods in three-dimensional space during computation. If the algorithm can find a non-overlapping arrangement such that all goods are contained within the effective space of the package, then... (True) indicates that the physical logic holds true; otherwise (False) indicates the existence of a physical conflict of "packing large goods in small boxes". The specific implementation logic of the 3D packing algorithm can be achieved by those skilled in the art using heuristic search algorithms or genetic algorithms, which are well-known technologies in this field and will not be elaborated upon here.

[0101] Step S302: Calculate the theoretical minimum remaining void ratio.

[0102] After passing the geometric verification ( Under the premise of ), the packing simulation unit 131 further calculates the theoretical void ratio inside the package. This indicator is used to quantify the utilization efficiency of packaging space, and the calculation formula is as follows:

[0103] ;

[0104] In the formula, the numerator is the theoretical natural cumulative volume of all goods in the order (without considering compression), and the denominator is the actual external volume of the logistics package. The range of values ​​is [0,1). The external volume is used as the denominator here to reflect the overall space occupied, including packaging materials.

[0105] Step S303: Perform an economic logic determination based on the freight-to-value ratio.

[0106] Packing simulation unit 131 combines logistics costs and the value of goods to construct an economic rationality verification model. In normal international trade, merchants tend to optimize packaging to reduce logistics costs, so it is rare for "freight costs to be much higher than the value of goods" and "the package to contain a lot of empty space" to occur, unless the transaction itself is intended to cover up some kind of fraudulent behavior (such as order brushing or tax fraud).

[0107] Define freight value ratio :

[0108] ;

[0109] In the formula, The freight amount is the actual freight cost in the logistics measurement vector, and the denominator is the total declared value of goods in the order transaction vector.

[0110] Define the economic logic verification result The judgment logic is as follows:

[0111] ;

[0112] In the formula, The preset porosity warning threshold, These are preset warning thresholds for the proportion of shipping costs. These two thresholds are dynamically set by the system based on historical big data. Specifically, the system statistically analyzes all normal transaction data for this category of goods over the past year and calculates the 90th quantile of the gap rate distribution as... The 95th percentile of the freight-to-value ratio distribution is calculated as... For example, in the general consumer electronics category, The value can be 0.5, but in the category of lightweight goods (such as pillows), the value can be adjusted to 0.7.

[0113] When the calculated porosity Exceeding the warning threshold, and the freight value is higher than When the threshold is also exceeded, the system determines that the transaction violates the logic of economic rationality (i.e., There is a risk that excessive packaging may conceal the actual absence of goods or that low-value goods may be used to fraudulently obtain export tax rebates; conversely, it is considered to be in line with economic logic. The judgment result is consistent with... Together, they constitute a dual verification conclusion in both the physical and economic dimensions.

[0114] See attached document Figure 2 Step S400 is executed by the density mapping unit 132 in the verification module 130. It aims to utilize the inherent physical density properties of substances to verify whether the declared product category matches the actual shipped physical entity, thereby identifying compliance risks such as "mismatched goods" or "falsely declared product names." This step is based on the physical axiom that "different substances have specific density distribution ranges." By comparing the theoretical mixed density with the actual apparent density, it determines whether there is a significant deviation in physical material between the package contents and the declared information. The density mapping unit 132 calls the HS code density mapping table constructed in step S100 and the logistics measurement vector generated in step S200, specifically including the following steps:

[0115] Step S401: Calculate the apparent density of the logistics parcels.

[0116] Density mapping unit 132 based on the measured vector of logistics The apparent density of the package is calculated based on the weight and dimensions. Apparent density differs from the absolute density of a material; it encompasses the combined mass distribution characteristics of the product itself, the internal packaging materials, and the voids.

[0117] The formula for calculating apparent density is defined as follows:

[0118] ;

[0119] In the formula, The total measured weight of the package is in kilograms. , , The actual measured length, width, and height dimensions of the package (unit: centimeters). This is a unit correction factor used to convert calculation results into standard density units (kg / m³). 3 In this embodiment, since the input unit is centimeters and the output needs to be converted to cubic meters, it is set as follows: =1,000,000.

[0120] In this process, if the logistics data source includes volumetric weight, the system will prioritize using the measured physical weight and measured geometric volume for calculation to reflect the true physical properties.

[0121] Step S402: Construct the dynamic mixing density feature range.

[0122] For common cross-border e-commerce orders involving "multiple items in one order" or "mixed products of various types," the density mapping unit 132 cannot simply directly call the density range of a single HS code. Based on the principles of mixture physics, the overall theoretical density of a system composed of multiple substances with different densities should be equal to the volume-weighted average of the densities of each component. Therefore, the unit needs to determine the density range based on the order transaction vector group. Based on the declaration information of each commodity, a composite fair density range is constructed for that specific order.

[0123] For including For orders of different types of goods, the system first calculates the value of each type of goods. Theoretical volume percentage :

[0124] ;

[0125] In the formula, For goods The number of applications, For goods The standard physical dimensions.

[0126] Subsequently, the system extracts the fair density range of the HS code corresponding to each product from the HS code density mapping table. and utilize volume ratio Perform weighted aggregation to generate the theoretical lower limit of the composite density for this order. and upper limit :

[0127] ;

[0128] ;

[0129] This yields the dynamic fair density range of the mixed package. This calculation logic ensures that the verification standard can dynamically adapt to the physical characteristics of different product combinations.

[0130] Step S403: Perform attribute consistency determination.

[0131] Density mapping unit 132 calculates the apparent density obtained in step S401. The dynamic fair density interval constructed in step S402 A comparison is performed. Considering potential systematic errors in logistics measurement equipment and fluctuations in packaging material weight, a tolerance buffer factor is introduced into the system. The judgment interval is corrected.

[0132] Define attribute consistency check results :

[0133] ;

[0134] In the formula, The value range is set to 0.05 to 0.10 (i.e., a deviation of 5% to 10% is allowed).

[0135] when When the physical attributes of the declared goods match the actual shipped items, the attribute consistency check is passed; when If an anomaly is detected, for example, if the declared commodity is "cotton fabric" (low density range), but the actual density falls within the range of "hardware" (high density range), the system will automatically identify the difference and mark it as a high-risk transaction.

[0136] See attached document Figure 2 Step S500 is executed by the quality and time series analysis unit 133 in the verification module 130. It aims to identify abnormal transactions such as "empty package order brushing" or "time reversal" by comparing the balance calculation of physical quality with the logical order of business occurrence time. This step is based on the fundamental axiom of "conservation of mass" and the irreversible causal logic of "order first, then delivery" in commercial activities, constructing a third verification barrier independent of geometric and economic dimensions. The time series analysis unit 133 calls the SKU physical attribute database constructed in step S100 and the vector data generated in step S200, specifically including the following steps:

[0137] Step S501: Construct a theoretical total weight model that includes packaging losses.

[0138] To accurately verify the law of conservation of mass, the system needs not only to calculate the cumulative net weight of the goods but also to theoretically estimate the weight of the logistics packaging materials. The verification unit reads the order transaction vector group. The net weight data of the goods, combined with the actual logistics measurement vector. Calculate the theoretical total weight based on the dimensional data. .

[0139] The formula for calculating the theoretical total weight is defined as follows:

[0140] ;

[0141] In the formula, For the first The declared quantity of each type of commodity, For the first Standard net weight per unit of this product (from the SKU physical attribute database). This is the estimated weight of the packaging materials.

[0142] In order to improve To improve estimation accuracy, the system employs a corrugated cardboard fitting algorithm based on surface area. Assuming the logistics package uses a standard corrugated cardboard box, its unfolded surface area is approximately equal to the package's outer surface area. The calculation is as follows:

[0143] ;

[0144] In the formula, , , The actual measured dimensions of the package (unit: meters); This is the areal density coefficient of corrugated cardboard, typically taken as 0.15 kg / m³. 2 Up to 0.25kg / m 2 (Corresponds to single-corrugated to double-corrugated cardboard); This is a filler correction factor used to compensate for the weight of auxiliary materials such as tape and bubble wrap. It is the system default setting. .

[0145] Step S502: Perform mass conservation deviation determination.

[0146] The verification unit calculates the actual total weight of the logistics. With theoretical total weight relative deviation rate between :

[0147] ;

[0148] Define quality verification results The judgment logic is as follows:

[0149] ;

[0150] In the formula, This is the permissible quality deviation threshold. Considering weight fluctuations and the impact of humidity on carton weight in industrial production, the system dynamically sets this threshold based on weight range. :for Lightweight small items, Set to 0.15 (15%); for Standard parts, Set to 0.05 (5%).

[0151] when When, it indicates that the transaction conforms to the law of conservation of mass; when At that time, a warning will be issued regarding the risk of "empty packages" or "smuggled goods".

[0152] Step S503: Perform business sequence causality verification.

[0153] The verification unit verifies the chronological order of capital flow, information flow, and logistics along the timeline based on the natural laws of commercial circulation. It reads the capital time-series vector. Order generation time Payment time and logistics pickup time .

[0154] Define timing verification results It must simultaneously meet the following two conditions:

[0155] Condition 1: Logical order compliance. That is, the order generation must be earlier than or equal to the payment time, and the payment time must be earlier than or equal to the logistics pickup time (for pre-sale or cash on delivery models, the system can be configured to exempt the payment time order check, but the order time must be earlier than the logistics time).

[0156] ;

[0157] Condition 2: Business cycle compliance. This means that the time span between each node must be within a reasonable range to prevent duplicate declarations using outdated order data.

[0158] ;

[0159] In the formula, The maximum effective lifespan of an order is typically set at 30 days (i.e., 30 × 24 × 3600 seconds). If both of the above conditions are met, then... If any condition is not met (e.g., an error occurs), If the time reversal phenomenon occurs, or the time span exceeds a threshold, then... .

[0160] Finally, the verification unit will and The output is sent to the multidimensional fusion computing module for subsequent comprehensive risk scoring.

[0161] See attached document Figure 2 Step S600 is executed by the multi-dimensional fusion calculation unit 134 in the verification module 130, aiming to aggregate the independent verification results of the geometric, economic, density, quality, and time dimensions in the preceding steps into a single quantitative compliance index. This step is based on the principle of multi-source evidence mutual verification, eliminating potential misjudgments or omissions in single-dimensional verification through weighted fusion, thereby constructing a comprehensive portrait of trade authenticity. The multi-dimensional fusion calculation unit 134 receives the geometric verification results from step S300. Results of verification with economic logic The attribute consistency verification result from step S403 and the quality verification results from step S500. Timing verification results Specifically, it includes the following steps:

[0162] Step S601: Construct a weighted compliance trust model.

[0163] The multi-dimensional fusion calculation unit 134 assigns corresponding weight coefficients based on the importance of each verification dimension in determining the authenticity of trade. Considering that physical feasibility (geometry and quality) is a prerequisite for the existence of trade, the system first executes the "key indicator rejection" logic. , or If any item in the index is 0, the comprehensive compliance trust index will be directly and forcibly reduced. Weighted calculations will no longer be performed.

[0164] If all of the above key indicators are 1, then the system defines a comprehensive compliance trust index. The calculation formula is as follows:

[0165] ;

[0166] In the formula, , , , , The weight coefficients are respectively the weights for geometric space, economic logic, physical properties, mass conservation, and time series dimensions, and satisfy the normalization condition. In this embodiment, the weighting coefficients are set as follows: (geometry), (economy), (property), (quality), (Time series). This weighting reflects a risk control strategy that prioritizes physical verification over logical verification.

[0167] Step S602: Perform risk classification determination.

[0168] The multidimensional fusion computing unit 134 has a preset hierarchical threshold set. According to the calculation The transactions are divided into three risk levels:

[0169] ;

[0170] In the formula, The high confidence threshold is usually set to 0.95, which means that almost all validation items pass. The low confidence threshold is typically set to 0.80, indicating the presence of non-critical logical flaws (such as slightly lower economic efficiency). When the result is "Pass," the system generates a compliance verification pass instruction; when the result is "Review," the system marks the order as requiring manual review and highlights the specific dimensions that led to the deduction; when the result is "Reject," the system directly blocks the subsequent reporting process for this transaction data.

[0171] See attached document Figure 2 Step S700 is executed by the declaration generation module 150, aiming to solidify the verified original data and verification results into tamper-proof digital evidence and generate the final declaration document. In this step, the declaration generation module 150 utilizes the distributed ledger characteristics of blockchain by calling the underlying capabilities of the evidence storage module 140 to ensure that once the data is on the chain, it cannot be tampered with by any party (including the enterprise itself). Specifically, it includes the following steps:

[0172] Step S701: Generate a structured hash fingerprint.

[0173] The declaration generation module 150 will generate the original vector group generated in step S200. The comprehensive compliance trust index generated in step S600 The verification results of each item are transmitted to the evidence storage module 140. The evidence storage module 140 serializes and concatenates the above data to form the data packet to be stored. And use a cryptographic hash algorithm to generate a unique digital fingerprint for the data packet. .

[0174] To ensure the reproducibility of the hash value, the evidence storage module 140 uses a normalized JSON string conversion during serialization. Specifically, the hash calculation process is as follows:

[0175] First, the system performs normalized serialization processing on the order transaction vector group, logistics measurement vector, and fund time series vector respectively to ensure that the keys of all JSON objects are arranged in ascending order of ASCII code, so as to eliminate the influence of the randomness of the key-value pair order on the hash result;

[0176] Subsequently, the three serialized strings, the numerical string of the comprehensive compliance trust index, and the pre-generated random salt value are concatenated in a predetermined order to form a complete original data string to be encrypted.

[0177] Finally, the SHA256 secure hash algorithm is used to process the original data string, and a fixed-length hexadecimal string is output as the unique digital fingerprint of the transaction.

[0178] The random salt value, after generation, is bound to the transaction serial number and stored in a local secure database for subsequent verification. For the specific implementation of the SHA256 algorithm, those skilled in the art can use standard encryption algorithm libraries.

[0179] Step S702: Perform the smart contract on-chain operation.

[0180] The evidence storage module 140 invokes a pre-deployed evidence storage smart contract on the consortium blockchain. This smart contract contains a data writing function named uploadEvidence, which the evidence storage module 140 uses to write the generated hash fingerprint. Unique transaction serial number The current block timestamp is passed as an input parameter to this function. The smart contract executes the consensus mechanism, packages the above information into a new block, and returns the on-chain transaction hash (TxID). The notarization module 140 returns this TxID as a receipt for successful notarization to the declaration generation module 150.

[0181] Step S703: Construct digital compliance credentials.

[0182] The declaration generation module 150 receives the TxID returned from the blockchain, binds it with the original order information, and generates an export tax refund declaration detail form that conforms to the preset format standards of the tax authorities. At the same time, the module embeds a QR code or digital signature containing the TxID and a random salt value into the declaration form.

[0183] Regulatory agencies or auditors can scan the TxID in the credential to query the corresponding blockchain explorer. The system then recalculates the hash value based on the original data to be verified and the locally provided Salt value, and compares them. If they match, it proves that the physical logistics data, capital flow data, and compliance verification results of the export tax refund transaction have not been tampered with by any party since their generation, thus achieving a closed loop of data consistency verification based on physical-digital mapping.

[0184] To make the objectives, technical solutions, and advantages of this invention clearer, the invention 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 merely illustrative and not intended to limit the invention.

[0185] Example:

[0186] This example demonstrates the compliant tax refund application process for a cross-border e-commerce company exporting a batch of mixed goods (Bluetooth speakers and pure cotton T-shirts).

[0187] The processing flow of this embodiment is as follows:

[0188] Step S1: Benchmark Library Construction and Data Collection

[0189] The system first pre-sets the physical parameters of customs codes “85182100” (single speaker) and “61091000” (cotton T-shirt) in the benchmark library management module.

[0190] Speaker specifications: Physical vectors are [15cm, 8cm, 6cm, 0.4kg], and the fair density range is [0.5, 0.7] g / cm³. 3 The material properties are hard (deformation tolerance coefficient 1.0).

[0191] T-shirt data: Physical vector is [30cm, 25cm, 2cm, 0.2kg] (folded state), fair density range is [0.1, 0.15] g / cm³. 3 The material is soft (deformation tolerance coefficient 1.2).

[0192] When a cross-border transaction occurs, the data collection module captures the following data:

[0193] Order transaction data: Includes 2 speakers and 2 T-shirts, with a declared total price of $60.

[0194] Actual logistics measurement data: The logistics service provider reported that the actual measured dimensions of the package were 32cm × 22cm × 12cm (volume 8448cm²). 3 The actual total weight was 1.45kg, and the shipping cost was 40 RMB.

[0195] Funds timeline data: Order generation time T1, payment time T2, pickup and scan time T3.

[0196] Step S2: Perform multidimensional cross-validation

[0197] Reference Figure 3 The verification module performs the following logical verification on the above data:

[0198] 1. Physical space logic verification (geometric packing simulation):

[0199] The system virtually fits the geometric shapes of two speakers and two T-shirts into a 32cm×22cm×12cm package space.

[0200] Calculate the total volume of the goods: 2×(15×8×6)+2×(30×25×2 / 1.2 after compaction)≈1440+2500=3940cm³ 3 .

[0201] The theoretical minimum remaining porosity is calculated as: (8448-3940) / 8448≈53%.

[0202] If the preset warning threshold is 80% (i.e., empty box or extremely unloaded), the current 53% is within a reasonable range, and it is determined that there is no physical conflict.

[0203] Economic rationality check: Calculate the freight-to-value ratio. 40 RMB / (60 USD × 7.2) ≈ 9.2%. If the threshold is 50%, then this ratio is economically rational (no money laundering characteristics of low value and high freight are observed).

[0204] 2. Attribute consistency verification (density distribution matching):

[0205] Calculated apparent density: 1.45 kg / 8448 cm³ 3 ≈0.171g / cm 3 .

[0206] Generate dynamic fair density ranges: weighted according to the volume percentage of the product. Assume the speaker's volume percentage is approximately 36%, and the T-shirt's volume percentage is approximately 64%.

[0207] The theoretical central value of the mixed density is approximately 0.36 × 0.6 (speaker average) + 0.64 × 0.125 (T-shirt average) ≈ 0.216 + 0.08 = 0.296 g / cm³ 3 .

[0208] Considering the extremely low density of packaging fillers (foam, bubble wrap), which would lower the overall density, the system, after adjusting for the filler coefficient, generates a dynamic range with a lower limit of 0.15 g / cm³. 3 .

[0209] Judgment: The measured apparent density is 0.171 g / cm³. 3 It falls within the corrected dynamic fair density range, so the verification is successful.

[0210] 3. Joint logic verification (quality conservation):

[0211] Total net weight of goods: 2×0.4+2×0.2=1.2kg.

[0212] Estimate the weight of the packaging materials: Based on the surface area of ​​the package (32×22+32×12+22×12)×2=2704cm² 2 Multiply by the corrugated cardboard surface density coefficient of 0.08 g / cm³ 2 The estimated weight of the cardboard box is approximately 0.21 kg.

[0213] Theoretical total mass: 1.2 + 0.21 = 1.41 kg.

[0214] Deviation verification: |1.45(measured) - 1.41(theoretical)| = 0.04 kg. The deviation is minimal, consistent with the law of conservation of mass.

[0215] Step S3: Risk Assessment and Evidence Preservation

[0216] The multi-dimensional fusion computing unit summarized the above results and found no abnormal key indicators. The overall compliance and credit score was calculated to be 95 points (high credit). The system judged the transaction as "low risk".

[0217] Step S4: Application Generation

[0218] The evidence storage module serializes and concatenates the transaction ID, verification log, and reputation value, adds a random salt value "Salt_X9d", performs a SHA-256 hash operation, and generates a digital fingerprint "0x7f8a...". This fingerprint is written to the consortium blockchain, returning the transaction hash "0x9b2c...".

[0219] The declaration generation module generates an export tax refund declaration form and embeds a QR code containing "0x9b2c..." in the remarks column for tax authorities to scan and verify.

[0220] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A blockchain-based cross-border e-commerce compliance tax refund method, characterized in that, The method comprises the following steps: constructing a reference library containing commodity physical attribute parameters and fair density intervals of customs codes; in response to a cross-border transaction event, obtaining order transaction data, logistics measured data, and fund time series data; based on the reference library, performing multi-dimensional cross-validation on the order transaction data, logistics measured data, and fund time series data; the multi-dimensional cross-validation at least includes: physical space logical verification based on geometric packing simulation, attribute consistency verification based on density distribution matching, and joint logical verification based on mass conservation and time sequence; comprehensive the verification results of the multi-dimensional cross-validation, determine the risk judgment level of the transaction; in response to the transaction data passing the risk judgment, generate a digital fingerprint containing the verification data and write it into a distributed ledger network, obtain an on-chain notarization index to generate an export tax rebate declaration file with the index; the order transaction data contains commodity SKU list, declared quantity, declared unit price, and declared customs code; the logistics measured data contains the length, width, height of the logistics package measured, the total weight of the package measured, and the actual payment of the logistics freight amount; the physical space logical verification specifically includes: virtually packing all the commodities in the order into the logistics package space to determine whether there is a physical conflict; if there is no physical conflict, calculate the theoretical minimum residual void fraction of the package; calculate the total declared value of the commodities according to the order transaction data, and calculate the ratio of the logistics freight amount to the total declared value of the commodities to obtain the freight value ratio; when the theoretical minimum residual void fraction exceeds the preset void fraction warning threshold, and the freight value ratio also exceeds the preset freight proportion warning threshold, it is determined that the transaction violates the economic rationality logic; the attribute consistency verification specifically includes: calculate the apparent density of the package according to the total weight of the package measured and the length, width, and height of the logistics package measured in the logistics measured data; when the order transaction data contains multiple different commodities, calculate the theoretical volume proportion of each commodity according to the declared quantity and the standard physical size of each commodity, and use the theoretical volume proportion to aggregate the fair density interval of the corresponding customs code of each commodity to generate a dynamic fair density interval; determine whether the apparent density falls within the fair density interval in the reference library or the dynamic fair density interval.

2. The blockchain-based cross-border e-commerce compliance tax refund method according to claim 1, characterized in that, The construction of the reference library containing commodity physical attribute parameters and fair density intervals of customs codes comprises: establishing a mapping relationship between commodity SKU and physical vector, the physical vector containing length, width, height, and mass parameters; selecting historical transaction data of a specific customs code category within a preset time period as a sample set, calculating the ratio of the total weight of the logistics measured to the logistics measured packaging size of each sample in the sample set to obtain the historical apparent density; calculate the mean and standard deviation of the density of the sample set, and determine the fair density interval of the customs code category based on the mean and standard deviation. 3.The blockchain-based cross-border e-commerce compliance tax refund method of claim 1, wherein, the joint logical verification specifically includes: According to the declared quantity in the order transaction data and the single-piece standard net weight in the reference library, the cumulative net weight of the commodity is calculated, and according to the length, width, height and preset packing material area density coefficient of the logistics package measured in the logistics measured data, the estimated weight of the packing material is calculated; The cumulative net weight of the commodity and the estimated weight of the packing material are added to obtain a theoretical total mass, and it is verified whether the absolute deviation of the theoretical total mass and the measured total weight of the package in the logistics measured data is less than a preset tolerance; It is verified whether the time sequence logic and interval period of the order generation time, payment completion time and logistics collection scanning time in the fund time sequence data meet the preset rules.

4. The blockchain-based cross-border e-commerce compliance tax refund method according to claim 1, characterized in that, The comprehensive multi-dimensional cross-verification results are used to determine the risk judgment level of the transaction, including: Determine whether there is a key indicator judged as abnormal in the verification results of the physical space logic verification, the attribute consistency verification and the joint logic verification. If there is, directly set the comprehensive compliance reputation value to zero; If there is no key indicator judged as abnormal, respectively assign a corresponding weight coefficient to each verification result and perform weighted summation to obtain a comprehensive compliance reputation value; According to the preset classification threshold interval to which the comprehensive compliance reputation value belongs, the risk judgment level of the transaction is determined.

5. The blockchain-based cross-border e-commerce compliance tax refund method according to claim 4, characterized in that, The generation of the digital fingerprint containing the verification data and the writing into the distributed ledger network includes: The order transaction data, the logistics measured data and the fund time sequence data are normalized and serialized to ensure that the keys of the objects are arranged in a predetermined order; The serialized data string is spliced with the numerical string of the comprehensive compliance reputation value and the pre-generated random salt value; A hash algorithm is used to operate on the spliced data string to generate a digital fingerprint; The data writing function of the smart contract is called to write the digital fingerprint, the transaction unique serial number and the current block timestamp into the distributed ledger network.

6. The blockchain-based cross-border e-commerce compliance tax refund method according to claim 5, characterized in that, The chain storage index is obtained to generate an export tax refund declaration file with the index, including: The chain transaction hash returned by the smart contract is received as the chain storage index; The chain transaction hash is bound with the order transaction data to generate an export tax refund declaration detail table, and a digital signature or a two-dimensional code containing the chain transaction hash and the random salt value is embedded in the export tax refund declaration detail table.

7. The blockchain-based cross-border e-commerce compliance tax refund method according to claim 1, characterized in that, The physical space logic verification uses a deformation tolerance coefficient, which is used to compensate for the volume change of soft packing material under pressure in the geometric packing simulation. The deformation tolerance coefficient is defined based on the material category of the commodity.

8. The blockchain evidence-based cross-border e-commerce compliant tax refund intelligent filing system for performing the blockchain evidence-based cross-border e-commerce compliant tax refund method according to any one of claims 1-7, characterized in that, It includes: A reference library management module configured to build a reference library containing commodity physical attribute parameters and customs code fair density intervals; A data collection module configured to obtain order transaction data, logistics measured data and fund time sequence data in response to cross-border transaction events; The verification module is configured to perform multi-dimensional cross verification on the order transaction data, the logistics measured data and the fund time sequence data based on the benchmark library, and the multi-dimensional cross verification at least includes: physical space logical verification based on geometric packing simulation, attribute consistency verification based on density distribution matching, and joint logical verification based on mass conservation and time sequence; the verification module further includes a multi-dimensional fusion calculation unit, which is configured to comprehensively integrate the verification results of the multi-dimensional cross verification, calculate a comprehensive compliance credit value by using a weighted model, and determine a risk judgment level of the transaction according to the credit value; The evidence storage module is configured to receive the verification results and key parameters output by the verification module, generate a digital fingerprint containing the verification data and write it into a distributed ledger network; The declaration generation module is configured to respond to the transaction data passed by the risk judgment, call the evidence storage module to obtain the on-chain evidence index, and generate an export tax rebate declaration file based on the on-chain evidence index.

Images