A data center-based electronic voucher full life cycle auditing and tracing method

Abstract

The application relates to the fields of data processing and financial information technology, and discloses a kind of based on data center's electronic voucher full life cycle audit and traceability method, comprising: data processing unit extracts voucher circulation approval state and multi-level quota allocation topology matrix and is translated into abstract feature tuple, constructs life cycle circulation topology directed graph in memory, compares the graph with historical topology data in global circulation state pool to calculate topology collision coefficient, interrupts download thread when exceeding, serializes and compresses life cycle circulation topology directed graph when not exceeding, and weaves with file binary stream, generates self-contained audit evidence data package and outputs, the application can effectively block illegal circulation of non-real assets, so that the audit file realizes in-situ verification of integrity and authenticity in offline state.

Description

Technical Field

[0001] This invention belongs to the field of data processing and financial information technology, and in particular relates to a method for full lifecycle auditing and traceability of electronic vouchers based on a data center. Background Technology

[0002] In the current large-scale factoring business network operation environment, electronic vouchers, as the core asset carrier, form the basis for ensuring system security through their circulation audit and traceability mechanisms. Typically, the industry generally adopts a centralized storage architecture. At the initial review and re-review nodes of voucher issuance, the system automatically associates the corresponding trade background contracts and invoice documents through primary and foreign key association rules. When responding to the data download request of the auditor, it directly packages the retrieved associated files. Under normal working conditions, this mode can only provide basic static file aggregation capabilities. However, this traditional association method relies entirely on the static indexing mechanism of relational databases. At the underlying logic level, it processes background information as a stateless data snapshot. As the scale of concurrent circulation increases, the static snapshot severs the spatiotemporal causal relationship between background files at circulation nodes such as customer access, credit allocation, initial voucher review, voucher re-review, and settlement and payment. The system cannot intercept the entry of non-genuine business data in real time, resulting in the malicious mounting of the same invoice characteristics across multiple batches of vouchers, thus causing semantic distortion.

[0003] While the aforementioned centralized storage architecture has undergone some optimization in terms of data throughput and physical topology at the hardware level, these infrastructure-level improvements have not addressed the core of business logic control. Furthermore, the flow control methods and data verification mechanisms at the upper software level also have significant shortcomings. For example, Chinese invention patent application CN121388052A discloses an AI-based electronic invoice lifecycle management system and method. This system constructs a link diagram including invoices, entities, and devices, and uses frequency and proportion thresholds to identify abnormal links. However, the underlying objective attributes upon which this technical solution implicitly relies are... This is a flat, linear relationship between the parties to a bill transaction and the operating equipment. However, when faced with the complex dynamic multi-level credit limit splitting and spatiotemporal causal evolution throughout the entire life cycle in large-scale factoring business, this control method, which relies on superficial statistical features, exposes its fundamental limitations. Its link diagram lacks causal chain modeling based on deep control levels such as the access compliance vector and the topology matrix of the credit limit allocation tree. It cannot perceive the deep causal distortions caused by the evolution of the credit limit distribution ratio between nodes. This makes it prone to serious anti-conflict circuit breaker omissions when dealing with highly concealed unauthorized transfer scenarios such as malicious reuse of background information across vouchers due to mismatch of basic constraints.

[0004] Therefore, the technical problem to be solved by this invention is how to reconstruct the interactive coupling mode of the control parameters of the entire life cycle of electronic vouchers from the perspective of data flow control, utilize the state characteristics of multiple modules to real-time topologically construct a dynamic flow topology network with strong causal chains in the system memory space, and improve the anti-conflict response performance of the system under multi-factor coupling conditions by using a highly cohesive serialization encapsulation mechanism based on the offline self-verification capability of the exported evidence storage byte stream. Summary of the Invention

[0005] This invention aims to solve the problem that traditional electronic voucher management systems rely entirely on static relational indexes, which sever the spatiotemporal causal chain of background information across all lifecycle nodes. This results in the system being unable to identify malicious reuse of background information across vouchers before packaging and downloading, and the exported data packets lacking offline in-situ self-verification capabilities.

[0006] In this technical solution, a method for full lifecycle auditing and traceability of electronic vouchers based on a data center includes the following steps: Step S1: The data processing unit extracts the full lifecycle data corresponding to the electronic voucher from the memory data structure and translates it into an abstract feature tuple; the full lifecycle data includes the admission status vector, the quota allocation tree topology matrix, and the circulation approval status. Step S2: The data processing unit constructs a directed acyclic graph of the lifecycle flow topology corresponding to the electronic voucher in memory, using the admission status vector as the source point, the evolution path of the quota allocation tree topology matrix as the directed edge, and the flow approval status as the sink point to establish a flow topology association network. Step S3: The data processing unit compares and arbitrates the directed acyclic graph of the lifecycle flow topology with the historical topology data in the global flow state pool, and calculates the topology collision coefficient of the trade background agreement file associated with the electronic voucher; when the topology collision coefficient is greater than or equal to a set threshold, the download thread of the trade background agreement file is interrupted and a logical inconsistency signal is output; when the topology collision coefficient is less than the set threshold, the directed acyclic graph of the lifecycle flow topology is compressed and a serialized topology path feature vector and an in-situ verification logic operator are generated. By rewriting the binary file header of the trade background agreement file, the serialized topology path feature vector is weaved and encapsulated with the binary stream of the trade background agreement file, and the in-situ verification logic operator is written into the reserved area of ​​the binary stream, generating a self-contained audit evidence storage data package and outputting it.

[0007] Preferably, in step S1, extracting the full lifecycle data corresponding to the electronic certificate includes: authenticating the applicant's corporate identity and assessing the real-name compliance status through the corporate customer module, generating an access status vector to indicate the access compliance result; retrieving the corresponding multi-level credit allocation tree topology distributed by the core enterprise through the credit limit management module, calculating the directed distribution ratio and evolution path of credit limits among multi-level nodes, and generating a credit limit allocation tree topology matrix to characterize the credit limit control level restrictions.

[0008] Preferably, in step S2, a directed acyclic graph of the lifecycle flow topology corresponding to the electronic voucher is constructed in memory, including the following sub-steps: Step S21, extracting the data download records of the electronic voucher in the preliminary review stage and the review stage, and obtaining the historical operator's packaged download operation record data of the trade background agreement document; Step S22, converting the packaged download operation record data into status nodes on the corresponding time axis, and attaching the status nodes to the corresponding topology branch of the directed acyclic graph of the lifecycle flow topology, forming a traceability chain covering the flow trajectory of the trade background agreement document.

[0009] Preferably, in step S3, the comparison and arbitration of the directed acyclic graph of the lifecycle flow topology with the historical topology data in the global flow state pool includes the following sub-steps: Step S31, calling the built-in topology collision coincidence verification algorithm operator to scan and compare the graph isomorphism and node collision of the currently constructed directed acyclic graph of the lifecycle flow topology with the historical topology data in the global flow state pool; Step S32, establishing a preset audit time window, extracting the multi-dimensional key feature components of the trade background agreement document associated with the current voucher, calculating the coincidence probability of the multi-dimensional key feature components within the preset audit time window, and determining whether the trade background agreement document has unauthorized duplicate flow across vouchers; Step S33, when the topology collision coefficient is greater than or equal to a set threshold, outputting a control command to stop the current data transmission, completing the thread-level blocking of the data packaging and download thread, and stopping the unauthorized flow topology associated network data flow.

[0010] Preferably, in step S3, performing serialization compression on the lifecycle flow topology directed acyclic graph and generating serialized topology path feature vectors and in-situ verification logic operators includes: using a graph structure serialization algorithm to extract the topology path of the lifecycle flow topology directed acyclic graph and compressing the binary state to generate a fixed-length serialized topology path feature vector; configuring an offline self-decryption verification algorithm program that can run independently of the factoring operation system network environment, and encapsulating it as an in-situ verification logic operator into a self-contained audit and evidence storage data packet.

[0011] Preferably, in step S3, the serialized topology path feature vector is weaved and encapsulated with the binary stream of the trade background agreement file by rewriting the binary file header of the trade background agreement file, including: locating the extended metadata file header area of ​​the trade background agreement file, modifying the length marker and check bit offset parameters of the extended metadata file header; converting the serialized topology path feature vector into a patch binary stream, and interleaving the patch binary stream into the blank reserved area of ​​the binary stream of the trade background agreement file, thereby completing the self-contained hybrid embedding of the serialized topology path feature vector with the original contract and invoice files.

[0012] Preferably, the following control steps are also included: Step S4, continuously track and record the state flow trajectory of the electronic voucher throughout its entire life cycle along the time axis, and generate a time series state dataset; Step S5, calculate the state evolution fluctuation index characterizing the stability of voucher circulation based on the time series state dataset, the state evolution fluctuation index including the rate of change and the degree of deviation; Step S6, configure a fixed circulation security threshold boundary, and when the state evolution fluctuation index continuously exceeds the circulation security threshold boundary, output predictive blocking information and automatically switch the circulation approval security control level of the electronic voucher.

[0013] Preferably, when performing offline verification of the in-situ verification logic operator, the upper limit of the kernel response allocation time of the data processing unit is set to 10μs; and on the premise of generating the time series state dataset in step S4, traffic data under the baseline working condition is collected, and the variation range of the overall feature interaction data is limited to within 0.8 to 1.2 by adjusting the thread bandwidth parameter.

[0014] Preferably, after generating and outputting the self-contained audit evidence storage data package, the following background evidence storage steps are also included: Step S7, the self-contained audit evidence storage data package is synchronously backed up to the local centralized audit history database, the incremental traceability log is written, and a globally unique index identifier for the self-contained audit evidence storage data package is generated so that the audit control terminal can trace the offline full flow trajectory based on the globally unique index identifier.

[0015] Compared with existing technologies, the present invention provides a method for full lifecycle auditing and traceability of electronic vouchers based on a data center, which has the following advantages: 1. In the full lifecycle audit of electronic vouchers, the system's central processing unit extracts the access status vector corresponding to the enterprise customer module, the quota allocation tree topology matrix corresponding to the quota management module, and the circulation approval status corresponding to the voucher issuance module. In real time, a directed acyclic graph data structure with a lifecycle circulation topology uniquely corresponding to the electronic voucher is constructed in the system's memory space. By using the feature tuples corresponding to the translated access real-name authentication information as the source point of the graph, the quota allocation evolution path as the directed edge, and the voucher approval status as the target sink point, a closed-loop network with spatiotemporal evolution characteristics is established. This eliminates the defect of traditional relational primary and foreign key indexes that sever node connections and realizes deep collaborative verification of multi-module management and control parameters.

[0016] 2. The system calls the built-in topology collision check operator to complete the topology state arbitration between the current directed acyclic graph of the system and the historical topology data in the global flow state pool. By calculating the sum of the number of topology nodes and directed edges that are determined to be overlapping and the number of legally shared feature tuples, and dividing it by the total number of feature tuples generated by the background protocol file, the topology collision coefficient reflecting the data coordination consistency is obtained. Under the triggered condition, the central processing unit shuts down the current data packaging and download thread and outputs a logical inconsistency signal to the console to prevent the malicious reuse and flow of assets across vouchers caused by the entry of non-real business data.

[0017] 3. When the topology verification passes, the processor performs serialization compression on the directed acyclic graph of the lifecycle flow topology, generating a serialized topology path feature vector and an in-situ verification logic operator. Then, by rewriting the binary file header of the protocol file, the serialized topology path feature vector is mixed and encapsulated with the binary stream of protocol files such as contracts and invoices to generate the final self-contained audit evidence data packet and output it to the download end. This allows the exported audit file to be removed from the factoring operation system network environment and still be able to complete the in-situ verification of the integrity and authenticity of the entire lifecycle by relying on its own embedded operators in an offline state, eliminating the inherent dependence on the centralized live database. Attached Figure Description

[0018] Figure 1 This is a flowchart of the topology comparison and evidence storage data encapsulation generation process of the present invention; Figure 2 This is a diagram of the terminal interaction and offline trajectory verification architecture of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0020] A method for full lifecycle auditing and traceability of electronic vouchers based on a data center includes the following steps: Step S1: The data processing unit extracts the full lifecycle data corresponding to the electronic voucher from the memory data structure and translates it into an abstract feature tuple; the full lifecycle data includes the admission status vector, the quota allocation tree topology matrix, and the circulation approval status. Step S2: The data processing unit constructs a directed acyclic graph of the lifecycle flow topology corresponding to the electronic voucher in memory, using the admission status vector as the source point, the evolution path of the quota allocation tree topology matrix as the directed edge, and the flow approval status as the sink point to establish a flow topology association network. Step S3: The data processing unit compares and arbitrates the directed acyclic graph of the lifecycle flow topology with the historical topology data in the global flow state pool, and calculates the topology collision coefficient of the trade background agreement file associated with the electronic voucher; when the topology collision coefficient is greater than or equal to a set threshold, the download thread of the trade background agreement file is interrupted and a logical inconsistency signal is output; when the topology collision coefficient is less than the set threshold, the directed acyclic graph of the lifecycle flow topology is compressed and a serialized topology path feature vector and an in-situ verification logic operator are generated. By rewriting the binary file header of the trade background agreement file, the serialized topology path feature vector is weaved and encapsulated with the binary stream of the trade background agreement file, and the in-situ verification logic operator is written into the reserved area of ​​the binary stream, generating a self-contained audit evidence storage data package and outputting it.

[0021] Preferably, in step S1, extracting the full lifecycle data corresponding to the electronic certificate includes: authenticating the applicant's corporate identity and assessing the real-name compliance status through the corporate customer module, generating an access status vector to indicate the access compliance result; retrieving the corresponding multi-level credit allocation tree topology distributed by the core enterprise through the credit limit management module, calculating the directed distribution ratio and evolution path of credit limits among multi-level nodes, and generating a credit limit allocation tree topology matrix to characterize the credit limit control level restrictions.

[0022] Preferably, in step S2, a directed acyclic graph of the lifecycle flow topology corresponding to the electronic voucher is constructed in memory, including the following sub-steps: Step S21, extracting the data download records of the electronic voucher in the preliminary review stage and the review stage, and obtaining the historical operator's packaged download operation record data of the trade background agreement document; Step S22, converting the packaged download operation record data into status nodes on the corresponding time axis, and attaching the status nodes to the corresponding topology branch of the directed acyclic graph of the lifecycle flow topology, forming a traceability chain covering the flow trajectory of the trade background agreement document.

[0023] Preferably, in step S3, the comparison and arbitration of the directed acyclic graph of the lifecycle flow topology with the historical topology data in the global flow state pool includes the following sub-steps: Step S31, calling the built-in topology collision coincidence verification algorithm operator to scan and compare the graph isomorphism and node collision of the currently constructed directed acyclic graph of the lifecycle flow topology with the historical topology data in the global flow state pool; Step S32, establishing a preset audit time window, extracting the multi-dimensional key feature components of the trade background agreement document associated with the current voucher, calculating the coincidence probability of the multi-dimensional key feature components within the preset audit time window, and determining whether the trade background agreement document has unauthorized duplicate flow across vouchers; Step S33, when the topology collision coefficient is greater than or equal to a set threshold, outputting a control command to stop the current data transmission, completing the thread-level blocking of the data packaging and download thread, and stopping the unauthorized flow topology associated network data flow.

[0024] Preferably, in step S3, performing serialization compression on the lifecycle flow topology directed acyclic graph and generating serialized topology path feature vectors and in-situ verification logic operators includes: using a graph structure serialization algorithm to extract the topology path of the lifecycle flow topology directed acyclic graph and compressing the binary state to generate a fixed-length serialized topology path feature vector; configuring an offline self-decryption verification algorithm program that can run independently of the factoring operation system network environment, and encapsulating it as an in-situ verification logic operator into a self-contained audit and evidence storage data packet.

[0025] Preferably, in step S3, the serialized topology path feature vector is weaved and encapsulated with the binary stream of the trade background agreement file by rewriting the binary file header of the trade background agreement file, including: locating the extended metadata file header area of ​​the trade background agreement file, modifying the length marker and check bit offset parameters of the extended metadata file header; converting the serialized topology path feature vector into a patch binary stream, and interleaving the patch binary stream into the blank reserved area of ​​the binary stream of the trade background agreement file, thereby completing the self-contained hybrid embedding of the serialized topology path feature vector with the original contract and invoice files.

[0026] Preferably, the following control steps are also included: Step S4, continuously track and record the state flow trajectory of the electronic voucher throughout its entire life cycle along the time axis, and generate a time series state dataset; Step S5, calculate the state evolution fluctuation index characterizing the stability of voucher circulation based on the time series state dataset, the state evolution fluctuation index including the rate of change and the degree of deviation; Step S6, configure a fixed circulation security threshold boundary, and when the state evolution fluctuation index continuously exceeds the circulation security threshold boundary, output predictive blocking information and automatically switch the circulation approval security control level of the electronic voucher.

[0027] Preferably, when performing offline verification of the in-situ verification logic operator, the upper limit of the kernel response allocation time of the data processing unit is set to 10μs; and on the premise of generating the time series state dataset in step S4, traffic data under the baseline working condition is collected, and the variation range of the overall feature interaction data is limited to within 0.8 to 1.2 by adjusting the thread bandwidth parameter.

[0028] Preferably, after generating and outputting the self-contained audit evidence storage data package, the following background evidence storage steps are also included: Step S7, the self-contained audit evidence storage data package is synchronously backed up to the local centralized audit history database, the incremental traceability log is written, and a globally unique index identifier for the self-contained audit evidence storage data package is generated so that the audit control terminal can trace the offline full flow trajectory based on the globally unique index identifier.

[0029] Example 1: When the current data center responds to the request to download the data package of electronic vouchers in the multi-party supply chain trade finance credit allocation business, the primary and foreign key static index structure of the relational database is prone to severing the spatiotemporal causal relationship between trade background agreement documents at different flow nodes. This makes it difficult to identify the repeated use of the same trade background agreement document among different electronic vouchers when downloading multiple vouchers in parallel. In the specific implementation process, the data processing unit extracts the full life cycle data corresponding to the electronic voucher from the memory data structure and translates the full life cycle data into abstract feature tuples.

[0030] During the translation process, the data processing unit first performs fixed-length field serialization on the extracted heterogeneous data. For the admission status vector, it extracts the compliance flag and real-name level number and converts them into 16-bit binary code. For the credit allocation tree topology matrix, it extracts the directed distribution ratio of credit limits among multi-level nodes and the parent-child node identifiers and converts them into a one-dimensional continuous fixed-point numerical sequence. For the approval status, it extracts the current node approval sequence number and review timestamp and converts them into time-series feature codes. Subsequently, the data processing unit concatenates the above 16-bit binary code, one-dimensional continuous fixed-point numerical sequence, and time-series feature codes according to a preset fixed offset to form a highly cohesive variable-length structure. The header of this structure contains a type identifier, and the tail contains a checksum, thereby uniformly translating the admission, credit limit, and approval information into abstract feature tuples that can be read and processed by the subsequent topology graph structure.

[0031] When constructing a directed acyclic graph (DAG) of the lifecycle flow topology in memory, the data processing unit uses the access status vector obtained through the enterprise customer module authentication as the source point of the graph, converts the evolution path of the quota allocation tree topology matrix extracted through the quota management module into directed edges, and uses the flow approval status corresponding to the voucher issuance module as the sink point. This establishes a flow topology association network. For the document download records of electronic vouchers in the initial review and re-review stages, the data processing unit reads the historical operator's packaged download operation record data of trade background agreement documents, and converts this packaged download operation record data into status nodes on the timeline according to the operation time, operator identifier, download batch identifier, and corresponding trade background agreement document identifier. The converted status nodes are then attached to the topology branches in the lifecycle flow topology DAG of the graph corresponding to the access status vector, quota allocation tree topology matrix, and flow approval status, according to their timestamp order. This ensures that the flow trajectory of the trade background agreement documents in the initial review, re-review, and packaged download processes falls into the same traceability chain.

[0032] The data processing unit calls the built-in topology collision and overlap verification algorithm operator to compare and arbitrate the currently constructed directed acyclic graph of lifecycle flow topology with the historical topology data stored in the global flow state pool. It then parses the multi-dimensional key feature components of the trade background agreement document associated with the current electronic voucher and calculates the topology collision coefficient. The calculation formula is as follows: ,in, The topological collision coefficient is... This refers to the number of topological nodes and directed edges in the global circulation state pool that are determined to overlap in terms of the characteristics of the current trade background agreement document within a preset time window. To determine the number of context feature tuples that are authorized for sharing within the current credit allocation tree whitelist. The total number of feature tuples generated for the current electronic voucher background file; superscript Indicates a collision, subscript Indicates overlap, subscript Indicates sharing, badge The total number is represented by the number of times the same file hash, the same invoice key feature component, and the same credit distribution node timestamp overlap in the historical topology data within a preset time window. The ratio of this number to the total number of historical feature components compared within the window is then used to obtain the overlap probability used to determine unauthorized duplicate transfers across vouchers. This overlap probability, together with the scanning results of graph isomorphism and node collision, serves as the basis for determining the topology collision coefficient before calculation.

[0033] When scanning for graph isomorphism and node collision, the topology collision overlap verification algorithm operator first obtains the total number of nodes and edges in the current lifecycle flow topology directed acyclic graph. Using this total number of nodes and edges as primary features, it performs a rapid coarse screening of the historical graph structure in the global flow state pool. For historical graphs that pass the coarse screening, the topology collision overlap verification algorithm operator uses a topology sequence sorting method based on node in-degree and out-degree to establish a one-to-one mapping candidate set for nodes in the current graph and historical graphs. Then, it compares the hash feature values ​​of the abstract feature tuples carried by each node layer by layer along the direction of the directed edges. When the node feature hashes of the corresponding positions on the same topology path in the two graphs are consistent, and the connection topology of the directed edges is consistent, it is determined that there is a graph isomorphism collision in the local network or the overall network, and the number of topology nodes and directed edges that are determined to be overlapping are accumulated and counted.

[0034] The data processing unit will convert the topology collision coefficients. Compared with the system's preset topology collision safety threshold, when the multidimensional key feature components of the trade background agreement document associated with the current electronic voucher extracted by the data processing unit overlap and match in the global flow state pool, resulting in a topology collision coefficient... When the error reaches 0.85 and the topology collision safety threshold is set to 0.70, the system determines that the current trade background agreement file has unauthorized duplicate circulation across documents. At this time, the data processing unit sends an interrupt signal to the memory control interface through the bus to terminate the current data packaging and download thread, and outputs a logical inconsistency signal to the console, thereby preventing the abnormal trade background agreement file from continuing to spread with the electronic document download process.

[0035] When the topological collision coefficient When the value is 0.35, which is below the topology collision safety threshold, the data processing unit uses the built-in graph structure serialization algorithm to serialize and compress the directed acyclic graph of the lifecycle flow topology, generate serialized topology path feature vectors and in-situ verification logic operators, and write the converted evidence storage byte stream into a specific offset address of the binary stream of the trade background agreement file, and outputs a self-contained audit evidence storage data packet containing offline in-situ self-verification data.

[0036] For the serialized topology path feature vector and the binary stream weaving and encapsulation of the trade background agreement document, the data processing unit first parses the standard file format header of the trade background agreement document, locates the extended metadata header area and the binary stream blank reserved area. Subsequently, the data processing unit adds the length value of the corresponding feature vector to the length marker field of the file and updates the cyclic redundancy check offset parameter of the global file stream. After completing the above processing, the data processing unit adopts a fixed 4-byte interleaved step size rule to convert the serialized topology path feature vector into a patched binary stream, and writes it into the binary stream blank reserved area non-destructively by interleaving 4 patch bytes every 1024 original data bytes. This weaving method maintains the normal reading logic of the original contract and invoice documents, and at the same time, during integrity verification, it can losslessly separate and extract the complete serialized topology path feature vector and in-situ verification logic operator according to the 4-byte interleaved step size rule.

[0037] During offline verification, the in-situ verification logic operator, which is an offline self-decryption verification algorithm program that can run independently of the factoring operation system network environment, is encapsulated in a self-contained audit and evidence storage data package. The offline verification end does not need to connect to the live database. It calls the offline self-decryption verification algorithm program embedded in the data package to read the serialized topology path feature vector woven in the binary file header and the binary stream blank reservation area. The program first extracts the asymmetric encryption public key that has been solidified at the time of asset issuance, decrypts the serialized topology path feature vector, and restores the complete life cycle flow topology directed acyclic graph data structure. Then, it reads the hash value of the trade background agreement file, the key feature components of the invoice, and the timestamps of the multi-level credit distribution nodes contained in the self-contained audit and evidence storage data package on the spot, and performs local causal chain closure verification with the restored graph structure. If the current physical attributes and business parameters of the file match the causal graph obtained by decryption, it is confirmed that the electronic certificate maintains authenticity and integrity in the offline state and reduces the dependence on real-time communication of the centralized global flow state pool.

[0038] Example 2: The current electronic voucher data processing method verifies its anti-counterfeiting and traceability effectiveness through a distributed financial data simulation test platform. This simulation test platform is equipped with a data processing unit with floating-point operation capabilities, a sampling frequency set to 50Hz, and a dynamic random access memory exchange bandwidth of no less than 10GB / s. It is used to reproduce the parallel voucher circulation scenario in multi-party supply chain financing business. The original input data used for the test is taken from a structured trade background dataset containing 50,000 characteristic records of actual factoring circulation business. To simulate the operating state of the data processing unit in a real engineering environment, a background disturbance component is superimposed on the input. This background disturbance component includes the injection of implicit negative event records caused by network latency when simulating high-concurrency access of the system. The injection frequency fluctuates randomly in the range of 1Hz to 5Hz.

[0039] In this simulation test platform, a 50Hz sampling frequency is used for statistical sampling of the overall traffic arrival status of the business layer, corresponding to an overall cycle of 20 milliseconds; the upper limit of kernel response allocation time of 10 microseconds is used for underlying hardware physical memory addressing and interrupt handling. The two time scales are asynchronously decoupled through a pointer circular queue with a double-buffered mirror structure. The business layer aggregates status data within a 20-millisecond cycle and pushes it in batches to the write cache of the pointer circular queue; the underlying kernel response runs independently with a 10-millisecond cycle, continuously searching the read cache and performing topology collision comparison, so that the overall sampling and the surface response remain continuous in the same data link.

[0040] In the experimental parameter settings, the topology collision security threshold was used as a key parameter affecting the accuracy of audit blocking. The value of this parameter is related to the historical topology base in the global circulation state pool and the preset time window span. When the topology collision security threshold is set too high, the filtering false negative rate of malicious reuse of trade background agreement documents across documents increases. When the topology collision security threshold is set too low, electronic documents in normal circulation are easily falsely deleted, thereby increasing the data processing load. The experiment determined the threshold by the cross-judgment relationship between the false positive rate and the false negative rate. Under the dense opening state where the average node growth rate within the preset time window is greater than 150 nodes per second, the topology collision security threshold approaches the lower limit of its value range. For the typical factoring financing situation corresponding to this experiment, the topology collision security threshold is set to 0.70.

[0041] The experiment sets up four structural states for comparison. In this embodiment, the admission state vector, the quota allocation tree topology matrix, and the transfer approval state are used to jointly establish a directed acyclic graph of the lifecycle transfer topology, and the topology collision overlap verification algorithm operator is activated. Its topology collision coefficient... Calculate according to the following formula: ,in, The topological collision coefficient is... This refers to the number of topological nodes and directed edges that are identified as overlapping in the global circulation state pool within a preset time window, representing the characteristics of the current trade background agreement document. To determine the number of context feature tuples that are authorized for sharing within the current credit allocation tree whitelist. The total number of feature tuples generated for the current electronic voucher background file; superscript Indicates a collision, subscript Indicates overlap, subscript Indicates sharing, badge Representing the total number, as a comparative example, keeping other features unchanged, removing the sink constraint of the approval status, relying only on static foreign keys for unidirectional association, the topology collision security threshold of the out-of-range control group 1 is reduced to 0.20, and the topology collision security threshold of the out-of-range control group 2 is increased to 0.90. The implicit injection ratio of background information being maliciously reused across vouchers is set to 10%, 35% and 85% respectively.

[0042] Under a 10% malicious injection ratio, the number of overlapping nodes detected in the sample group of this invention. The number is 11, and the number of shared features is 11. The total number of feature tuples is 16. The value is 114. Substituting the above data into the calculation relationship of the topological collision coefficient, we get... The value is 0.24; since 0.24 is lower than 0.70, the data processing unit does not trigger an interrupt, the data download thread is allowed to proceed normally, and a self-contained audit and evidence storage data packet is generated. Under a 35% malicious injection ratio, the number of overlapping nodes in the sample group of this invention is... Increased to 45, shared feature number The total number of feature tuples is 19. The value is 120, and the calculation yields... The value is 0.53; this value is still below 0.70, so the data download thread continues to be allowed to proceed normally, and a self-contained audit and evidence storage data packet is generated. Under an 85% malicious injection rate, the number of overlapping nodes is... Increased to 97, sharing feature number The total number of feature tuples is 15. The value is 132, calculated as follows: The value is 0.85; since 0.85 reaches and exceeds 0.70, the data processing unit triggers an interrupt signal, terminating the data download thread.

[0043] In the partially missing control group, under the same 10% malicious injection ratio, the removal of the sink constraint for the flow approval status prevented the formation of a complete lifecycle flow topology directed acyclic graph, resulting in a higher number of overlapping nodes. =0; in shared feature number The total number of characteristic tuples is 16. When the value is 114, the topological collision coefficient With a score of only 0.14, the interception function failed. Under a 35% malicious injection rate, the number of overlapping nodes in this control group was [missing data]. Still 0, shared feature number The total number of feature tuples is 19. The value is 120, and the calculation yields... With a value of 0.16, it is still impossible to construct a directed graph and complete the interception. At an 85% malicious injection rate, the number of overlapping nodes is [not specified]. Still 0, shared feature number The total number of feature tuples is 15. The value is 132, calculated as follows: The value was 0.11; even though the malicious reuse rate had risen to 85%, the control group still maintained a low output due to the lack of a settlement point constraint for the circulation approval status, and could not identify the cross-voucher reuse status.

[0044] In the out-of-range control group, after lowering the topology collision safety threshold to 0.20, the number of overlapping nodes was still used under a 10% malicious injection ratio. 11, shared feature number The total number of characteristic tuples is 16. Given 114 as input data, the calculation is as follows: The value is 0.24. Since 0.24 exceeds 0.20, this group will trigger an interruption even with low-level interference, causing the download thread to hang abnormally. With a 35% malicious injection rate, the number of overlapping nodes... The number is 45, and the number of shared features is 45. The total number of feature tuples is 19. It is 120. The value is 0.53; this value also exceeds 0.20, and the download thread continues to be abnormally suspended. With an 85% malicious injection rate, the number of overlapping nodes... The value is 97, and the number of shared features is... The total number of feature tuples is 15. It is 132. The threshold is 0.85; if the value exceeds 0.20, the data processing unit triggers an interrupt signal and terminates the data download thread. The above results indicate that if the threshold is too low, normal data download under low interference conditions will also be blocked.

[0045] In the out-of-range control group 2, after raising the topology collision safety threshold to 0.90, the number of overlapping nodes was reduced under a 10% malicious injection ratio. The number is 11, and the number of shared features is 11. The total number of feature tuples is 16. The value is 114, calculated as follows: The value is 0.24; since 0.24 is lower than 0.90, the data download thread is allowed to proceed normally and generates a self-contained audit and evidence storage data packet. Under a 35% malicious injection ratio, the number of overlapping nodes is... The number is 45, and the number of shared features is 45. The total number of feature tuples is 19. The value is 120, and the calculation yields... The value is 0.53; this value is still below 0.90, the data download thread is allowed to proceed normally and generates a self-contained audit and evidence storage data packet. Under an 85% malicious injection rate, the number of overlapping nodes is... The value is 97, and the number of shared features is... The total number of feature tuples is 15. The value is 132, calculated as follows: The value is 0.85; since 0.85 is still less than 0.90, the system does not trigger an interrupt, resulting in a missed report and the system is allowed to proceed.

[0046] As can be seen from the above data link, in the sample group of this invention, the malicious injection ratios of 10% and 35% correspond to topology collision coefficients of 0.24 and 0.53 respectively, both lower than 0.70. The data processing unit allows the data download thread and generates a self-contained audit and evidence storage data package; when the malicious injection ratio rises to 85%, the number of overlapping nodes... The topological collision coefficient increased further from 45 to 97. The value rose to 0.85, exceeding the topology collision safety threshold of 0.70. The data processing unit then locked the download thread in the data center via a bus interrupt command. In contrast, the partially missing control group, lacking time-series constraints between nodes throughout their entire lifecycle, had a higher number of overlapping nodes. At malicious injection rates of 10%, 35%, and 85%, the topology collision coefficients were all 0, with values ​​remaining at 0.14, 0.16, and 0.11 respectively. This failed to effectively block high-proportion malicious reuse. The out-of-range control group 1, at a threshold of 0.20, judged both 0.24 and 0.53 as interruption trigger states, with low and medium interferences causing abnormal suspensions. The out-of-range control group 2, at a threshold of 0.90, still judged 0.85 as not reaching the threshold, allowing extreme risk conditions to be allowed to proceed. Therefore, it can be seen that a topology collision safety threshold of 0.70 can distinguish between normal flow and unauthorized cross-credential duplicate flow under the conditions of this experiment.

[0047] In actual configuration, to ensure that the kernel response allocation time during offline verification is no more than 10 microseconds, the data processing unit utilizes dynamic random access memory with a bandwidth of no less than 10GB / s to dynamically exchange bandwidth. An interrupt-driven allocation mechanism is adopted in kernel mode. When the offline self-decryption verification algorithm is triggered, the data processing unit bypasses the context switching of the user-mode file system through preemptive kernel scheduling, completing the physical memory block addressing response within 10 microseconds. To control the overall feature interaction data variation within the range of 0.8 to 1.2, the simulation test platform monitors the frequency of incoming features in the cache queue in real time. When data fluctuations deviate from the baseline, a negative feedback adjustment loop is introduced by dynamically adjusting the thread's concurrent sliding window size and network bandwidth allocation weight. When traffic shows a sudden increase, the system proportionally tightens the thread bandwidth, constraining the overall feature interaction data frequency to within 0.8 to 1.2 times the baseline, keeping timing fluctuations within a controllable range during testing.

[0048] Example 3: In a high-concurrency settlement environment with a daily average data processing volume of more than 1,000,000 feature records, the historical topology data in the global flow state pool continues to accumulate. The directed acyclic graph of the lifecycle flow topology is prone to problems such as excessively long retrieval links and increased thread waiting as the scale of concurrent flow expands. In order to maintain the stability of memory retrieval latency, the data processing unit allocates a set of pointer circular queues with a dual-buffer mirror structure in memory. One side of the buffer receives the electronic voucher status data written by the business layer, and the other side of the buffer is used for the topology collision overlap verification algorithm operator to read continuously. The two buffers are switched alternately according to the queue pointer.

[0049] The data processing unit reads the admission state vector corresponding to the enterprise customer module and uses the multi-dimensional feature components in the admission state vector as hash index keys. It then maps the calculated feature space address to a dynamic storage block with continuous physical addressing characteristics. For the current transfer node timestamp passed from upstream, the data processing unit uses it as a time constraint condition for the topology node. For topology nodes that have not undergone credit transfer for more than 180 days, the unit performs head and tail pointer disconnection and memory space erasure, so that normal data blocks converge within the high-frequency retrieval envelope of 3 levels or less. In this way, low-frequency nodes in historical topology data no longer continuously occupy high-frequency retrieval paths. The data processing unit reduces feature physical collision misalignment and retrieval latency loss caused by data scale overload within the hardware clock cycle.

[0050] During continuous operation, the data processing unit tracks the state transition trajectory of the electronic voucher throughout its entire lifecycle along the timeline. It writes the admission state vector, the credit limit allocation tree topology matrix, the approval status, and the execution status of the data packaging and download thread into the time series state dataset by timestamp. This time series state dataset is used to record the continuous state changes of the electronic voucher from admission, credit limit distribution, approval flow to data download output. Subsequent state judgments are all based on the timestamps, state nodes, and topology edge changes in this dataset.

[0051] The data processing unit calculates a state evolution fluctuation index based on the time-series state dataset. This index includes the rate of change and the deviation. The rate of change is determined by the change in the number of state nodes or topological edges within adjacent sampling windows, while the deviation is determined by the degree of deviation of the state transition trajectory within the current window from the stable transition trajectory under the baseline conditions. The data processing unit compares the state evolution fluctuation index with a fixed transition security threshold boundary. When the rate of change and the deviation continuously exceed the transition security threshold boundary within consecutive sampling windows, the data processing unit outputs predictive blocking information to the audit control terminal and switches the security control level of the electronic voucher's transition approval to a higher level, so that subsequent data downloads, topological comparisons, and approval transitions all enter a more stringent verification path.

[0052] Based on the aforementioned state tracking, the data processing unit invokes the differential adaptive calibration operator to periodically adjust the safety boundary when the system determines the reuse status according to the transient arrival density of the current flow approval status. This differential adaptive calibration operator introduces a compensation operator. The corrected topology collision safety threshold is calculated as follows: ,in, For dimensionless compensation operators, This refers to the frequency of document arrival characteristics obtained by the data processing unit within the current 3-second sliding sampling window. The reference flow frequency constant is the fixed standard of the system, and the subscript is... Indicates the current item, subscript Indicates the benchmark term.

[0053] The data processing unit combines the topology collision safety threshold with the compensation operator. The product is used as a dynamic reduction boundary. When the distributed financial data simulation test platform experiences traffic aggregation due to high network concurrency latency disturbances, it causes... Reaching 1200 / s and When the value is fixed at 400 data points / second, the data processing unit obtains the compensation operator based on the above calculation relationship. The threshold is set to 3. Then, the initial topology collision safety threshold of 0.70 is multiplied by the compensation operator to obtain the adjusted temporary interception threshold of 2.10.

[0054] Under traffic aggregation conditions, the dynamic reduction boundary is increased to 2.10 due to the topology collision coefficient. The theoretical upper limit is 1. Even if the topology collision coefficient calculated at the current time reaches 1, it is still less than the dynamic reduction boundary. During the transient network delay disturbance of extremely high concurrency, the data processing unit enters a temporary tolerance release state. This state is used to reduce the risk of false killing of normal credentials caused by the instantaneous accumulation of feature frequencies. At the same time, the system still maintains the convergence of the underlying retrieval path through the continuous physical addressing of dynamic storage blocks, the disconnection of the first and last pointers, and the erasure of memory space, so that the concurrency risk no longer depends solely on the topology collision coefficient truncation processing.

[0055] If the currently calculated topology collision coefficient If the data is less than the dynamic reduction boundary, the data processing unit directly weaves the generated evidence storage byte stream into a specific offset of the binary stream of the corresponding trade background agreement file through the internal bus-level interface. Subsequently, the data processing unit uses the existing processor and dynamic storage block to periodically correct the traffic deviation caused by network disturbances through the data addressing feedback adjustment loop, so that the electronic voucher can still complete topology comparison, data download control, and self-contained audit evidence storage data packet output in a high-concurrency settlement environment.

[0056] Example 4: This example combines Figures 1 to 2 This paper describes a method for full lifecycle auditing and traceability of electronic vouchers based on a data center, such as... Figure 1 As shown, the entire lifecycle data is first extracted, including the admission status vector, the quota allocation tree topology matrix, and the transfer approval status. This data is then translated into abstract feature tuples to structure the entire lifecycle data. Next, a directed acyclic graph (DAG) of the lifecycle transfer topology is constructed, with the admission status as the source, the allocation path as the edge, and the transfer approval status as the sink. In the comparison and arbitration step, the DAG is compared with historical topology data provided and maintained by the global transfer status pool. The topology collision coefficient is then calculated to assess the probability of overlap in trade background agreement documents. It is then determined whether the topology collision coefficient is greater than or equal to a set threshold. If the result is yes (greater than or equal to the threshold), the download thread is interrupted and a logical inconsistency signal is output to block the illegal transfer of non-genuine assets. If the result is no (less than the threshold), feature vectors and verification operators are compressed and generated. Serialized topology path feature vectors and in-situ verification logic operators are generated by compressing the topology graph. Finally, a self-contained audit evidence data package is generated and output. This rewrites the binary file header and weaves and encapsulates the vectors and operators to achieve in-situ verification of integrity and authenticity.

[0057] like Figure 2 As shown, auditors interact with the request to download the packaged data, which triggers the extraction of full lifecycle data. The extracted full lifecycle data is input and translated into abstract feature tuples. Historical operators interact with the construction of a directed acyclic graph (DAG) of the lifecycle flow topology, and simultaneously restore missing nodes through a cyclic error correction area, inputting them into the constructed DAG of the lifecycle flow topology. The constructed DAG of the lifecycle flow topology points in reverse and depends on the translated abstract feature tuples. The completed DAG of the lifecycle flow topology further guides the arbitration of comparing historical topology data. The arbitration of comparing historical topology data points unidirectionally to the interrupted trade background agreement document download thread and to the offline full flow trajectory traced based on the globally unique index identifier. At the same time, a self-contained audit evidence data package is generated, pointing unidirectionally to and converging into the offline full flow trajectory traced based on the globally unique index identifier. The audit control terminal is connected to and acts on the offline full flow trajectory traced based on the globally unique index identifier, while the offline verification terminal is interconnected with the offline in-situ self-decryption verification, and this use case remains independent within the system.

[0058] Example 5: After the electronic voucher completes the data download control and generates a self-contained audit evidence storage data packet, the data processing unit synchronously writes the data packet into the local centralized audit history database. During the writing process, the local centralized audit history database uses the electronic voucher's business identifier, trade background agreement document identifier, and generation time as the retrieval entry point. It receives the serialized topology path feature vector, in-situ verification logic operator, trade background agreement document hash value, and corresponding flow status summary from the self-contained audit evidence storage data packet, and saves the above data as a complete audit evidence storage record.

[0059] To avoid the disconnect between offline verification results and central historical records, the data processing unit synchronously generates an incremental traceability log each time it writes a self-contained audit evidence storage data packet. This incremental traceability log records the state nodes, topological edges, file hash values, and data download control results that have changed in this evidence storage compared to the previous version. For historical topological data that has not changed, only its index back-pointing relationship is retained, and it is not repeatedly written to the complete data packet. In this way, the local centralized audit history database can retain the continuous flow and changes of electronic vouchers in chronological order, while reducing the storage occupation of duplicate historical data.

[0060] After generating the incremental traceability log, the data processing unit generates a globally unique index identifier for the self-contained audit evidence storage data package based on the electronic voucher identifier, the hash value of the trade background agreement document, the write timestamp, and the record sequence number of the local centralized audit historical database. This globally unique index identifier is bound and stored with the self-contained audit evidence storage data package, the incremental traceability log, and the corresponding historical topology data, so that the same electronic voucher forms a searchable index chain in each stage of access, credit limit distribution, approval process, document download, and offline verification.

[0061] After receiving the globally unique index identifier, the audit control terminal first queries the local centralized audit history database for the corresponding self-contained audit evidence data packet, and then reads the incremental traceability log bound to the data packet. The audit control terminal replays the changes of status nodes and topology edges in the order of timestamps in the log, and combines the serialized topology path feature vector to restore the circulation path of electronic vouchers throughout their entire life cycle. For data packets that have left the factoring operation system network environment, the audit control terminal can still trace the offline full circulation trajectory based on the globally unique index identifier, check whether the trade background agreement document is reused between different electronic vouchers, whether the data download thread is interrupted, and whether the offline verification results are consistent with the local historical records.

[0062] When the audit control terminal detects a data download thread interruption record in the incremental traceability log corresponding to a globally unique index identifier, it continues to read the topology collision coefficient, topology collision safety threshold, and logical inconsistency signal output status before and after the interruption record, and compares them with the in-situ verification results in the self-contained audit evidence storage data package. If both point to the unauthorized duplicate circulation of the same trade background agreement document across documents, the offline full circulation trajectory of the electronic document is marked as an abnormal trajectory. If the in-situ verification result is inconsistent with the incremental traceability log, the audit control terminal includes the data package corresponding to the globally unique index identifier in the manual review queue, retains its original evidence storage record, and does not overwrite the existing data in the local centralized audit historical database.

[0063] Through the above processing, the self-contained audit evidence storage data package is no longer just an isolated file after a single download, but forms a corresponding relationship with the incremental traceability logs, globally unique index identifiers, and historical topology data in the local centralized audit historical database. The audit control terminal does not need to re-access the real-time business database, but can retrieve the offline full flow trajectory based on the globally unique index identifier, and continuously trace the status of each stage of electronic vouchers from access to data download and output.

[0064] Example 6: During continuous downloading and offline verification, the data processing unit configures a 512KB cyclic error correction area for the electronic voucher's circulation topology network. The cyclic error correction area is written synchronously with the data download thread and is used to temporarily store the most recent valid record of the admission status vector, credit limit distribution tree topology matrix, circulation approval status, and trade background agreement document hash value. When network jitter or cache write-back delay causes signal packet loss, the data processing unit first reads the valid records in the adjacent time slice from the cyclic error correction area, and then restores the missing nodes and missing edges in the directed acyclic graph of the lifecycle circulation topology.

[0065] In an abnormal situation, the admission state vector received by the data download thread is misaligned with the credit limit distribution tree topology matrix. Instead of directly discarding the electronic voucher's transfer record, the data processing unit starts with the most recently valid admission state vector retained in the cyclic error correction area and continues matching along the still identifiable parent-child node identifiers in the credit limit distribution tree topology matrix. For topology edges that cannot be directly matched, a temporary connection is established according to the tail pointer of the previous node and the head pointer of the next node, so that the directed acyclic graph of the life cycle transfer topology remains traversable.

[0066] After the missing data recovery is completed, the data processing unit calculates the topology integrity index based on the number of recovered nodes and the expected number of nodes. In this recovery, the actual number of recovered record nodes in the cyclic error correction zone is 85. Compared with the expected number of nodes (100) that the data download thread should write within this time window, the topology integrity index is 0.85. This value is higher than the crash judgment threshold of 0.60. Based on this, the data processing unit retains the current lifecycle flow topology directed acyclic graph and continues to execute the generation of serialized topology path feature vectors. If the topology integrity index is lower than 0.60, the generation of this data packet is stopped, and the electronic certificate is marked as an abnormal flow record that needs to be re-collected.

[0067] For the topological edge formed by the temporary connection, the data processing unit reads the hash value of the trade background agreement file, the circulation approval status, and the output result of the in-situ verification logic operator again during subsequent verification. If the three are consistent with the most recent valid record in the cyclic error correction area, the temporary connection is solidified into the restored topological edge; if any result is inconsistent, the temporary connection is revoked, and the corresponding node is kept as an isolated node, waiting for the incremental traceability log in the local centralized audit historical database to fill in its source.

[0068] Under the aforementioned abnormal operating conditions, from the detection of signal packet loss, reading of valid records in the cyclic error correction area, completion of temporary topology edge connections, to the output of the restored lifecycle flow topology directed acyclic graph, the response time of the data processing unit does not exceed 12ms. The restored topology data continues to participate in the calculation of topology collision coefficients and offline verification, without changing the content of the original trade background agreement document already written in the self-contained audit evidence data packet, nor covering the existing audit evidence records in the local centralized audit historical database.

[0069] The embodiments of this application have been described above with reference to the accompanying drawings. Unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other. This application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit of this application and the scope of protection of this invention, and all of these forms are within the protection scope of this application.

Claims

1. A method for full lifecycle auditing and traceability of electronic vouchers based on a data center, characterized in that, Includes the following steps: Step S1: The data processing unit extracts the full lifecycle data corresponding to the electronic voucher from the memory data structure and translates it into an abstract feature tuple; the full lifecycle data includes the admission status vector, the quota allocation tree topology matrix, and the circulation approval status. Step S2: The data processing unit constructs a directed acyclic graph of the lifecycle flow topology corresponding to the electronic voucher in memory, using the admission status vector as the source point, the evolution path of the quota allocation tree topology matrix as the directed edge, and the flow approval status as the sink point to establish a flow topology association network. Step S3: The data processing unit compares and arbitrates the directed acyclic graph of the lifecycle flow topology with the historical topology data in the global flow state pool, and calculates the topology collision coefficient of the trade background agreement file associated with the electronic voucher; when the topology collision coefficient is greater than or equal to a set threshold, the download thread of the trade background agreement file is interrupted and a logical inconsistency signal is output; when the topology collision coefficient is less than the set threshold, the directed acyclic graph of the lifecycle flow topology is compressed and a serialized topology path feature vector and an in-situ verification logic operator are generated. By rewriting the binary file header of the trade background agreement file, the serialized topology path feature vector is weaved and encapsulated with the binary stream of the trade background agreement file, and the in-situ verification logic operator is written into the reserved area of ​​the binary stream, generating a self-contained audit evidence storage data package and outputting it.

2. The method for full lifecycle auditing and traceability of electronic vouchers based on a data center, as described in claim 1, is characterized in that... In step S1, the extraction of the full lifecycle data corresponding to the electronic certificate includes: authenticating the applicant's corporate identity and assessing the real-name compliance status through the enterprise customer module, generating an access status vector to indicate the access compliance result; retrieving the corresponding multi-level credit allocation tree topology distributed by the core enterprise through the credit limit management module, calculating the directed distribution ratio and evolution path of credit limits among multi-level nodes, and generating a credit limit allocation tree topology matrix to characterize the credit limit control level restrictions.

3. The method for full lifecycle auditing and traceability of electronic vouchers based on a data center as described in claim 1, characterized in that, In step S2, a directed acyclic graph of the lifecycle flow topology corresponding to the electronic voucher is constructed in memory, including the following sub-steps: Step S21, extract the data download records of the electronic voucher in the preliminary review stage and the review stage, and obtain the data of the historical operators' packaged download operation records of the trade background agreement documents; Step S22, convert the packaged download operation record data into the status nodes on the corresponding time axis, and attach the status nodes to the corresponding topology branches of the directed acyclic graph of the lifecycle flow topology to form a traceability chain covering the flow trajectory of the trade background agreement documents.

4. The method for full lifecycle auditing and traceability of electronic vouchers based on a data center as described in claim 1, characterized in that, In step S3, the lifecycle flow topology directed acyclic graph is compared and arbitrated with the historical topology data in the global flow state pool, including the following sub-steps: Step S31, the built-in topology collision coincidence verification algorithm operator is invoked to scan and compare the graph isomorphism and node collision of the currently constructed lifecycle flow topology directed acyclic graph with the historical topology data in the global flow state pool; Step S32, a preset audit time window is established, the multi-dimensional key feature components of the trade background agreement document associated with the current document are extracted, the coincidence probability of the multi-dimensional key feature components is calculated within the preset audit time window, and it is determined whether the trade background agreement document has unauthorized duplicate flow across documents; Step S33, when the topology collision coefficient is greater than or equal to the set threshold, a control command to stop the current data transmission is output, the thread-level blocking of the data packaging and download thread is completed, and the unauthorized flow topology associated network data flow is stopped.

5. The method for full lifecycle auditing and traceability of electronic vouchers based on a data center according to claim 1, characterized in that, In step S3, the serialization compression of the lifecycle flow topology directed acyclic graph and the generation of serialized topology path feature vectors and in-situ verification logic operators include: using a graph structure serialization algorithm to extract the topology path of the lifecycle flow topology directed acyclic graph and compressing the binary state to generate a fixed-length serialized topology path feature vector; configuring an offline self-decryption verification algorithm program that can run independently of the factoring operation system network environment, and encapsulating it as an in-situ verification logic operator into a self-contained audit and evidence storage data packet.

6. The method for full lifecycle auditing and traceability of electronic vouchers based on a data center as described in claim 1, characterized in that, In step S3, by rewriting the binary header of the trade background agreement file, the serialized topology path feature vector is weaved and encapsulated with the binary stream of the trade background agreement file, including: locating the extended metadata header area of ​​the trade background agreement file, modifying the length marker and check bit offset parameters of the extended metadata header; converting the serialized topology path feature vector into a patched binary stream, and interleaving the patched binary stream into the blank reserved area of ​​the binary stream of the trade background agreement file, thus completing the self-contained hybrid embedding of the serialized topology path feature vector with the original contract and invoice files.

7. The method for full lifecycle auditing and traceability of electronic vouchers based on a data center according to claim 1, characterized in that, The following control steps are also included: Step S4, continuously track and record the state flow trajectory of electronic vouchers throughout their entire lifecycle and generate a time series state dataset; Step S5, calculate the state evolution fluctuation index, which characterizes the stability of voucher circulation, based on the time series state dataset. The state evolution fluctuation index includes the rate of change and the degree of deviation; Step S6, configure a fixed circulation security threshold boundary, and when the state evolution fluctuation index continuously exceeds the circulation security threshold boundary, output predictive blocking information and automatically switch the circulation approval security control level of electronic vouchers.

8. The method for full lifecycle auditing and traceability of electronic vouchers based on a data center according to claim 7, characterized in that, When performing offline verification of the in-situ verification logic operator, the upper limit of the kernel response allocation time of the data processing unit is set to 10μs; and on the premise of generating the time series state dataset in step S4, traffic data under the baseline working condition is collected, and the variation range of the overall feature interaction data is limited to within 0.8 to 1.2 by adjusting the thread bandwidth parameter.

9. The method for full lifecycle auditing and traceability of electronic vouchers based on a data center as described in claim 1, characterized in that, After generating and outputting the self-contained audit evidence data package, the following background evidence storage steps are also included: Step S7, the self-contained audit evidence data package is synchronously backed up to the local centralized audit history database, the incremental traceability log is written, and a globally unique index identifier is generated for the self-contained audit evidence data package so that the audit control terminal can trace the offline full flow trajectory based on the globally unique index identifier.

Images