A smart contract-based anti-counterfeiting information verification system for gift certificates

Abstract

This invention relates to the field of blockchain and smart contract application technology, specifically a smart contract-based anti-counterfeiting information verification system for commodity vouchers, comprising: a request receiving module for receiving commodity voucher verification requests sent by terminals; a network status monitoring module for acquiring node computational status data and consensus delay data, and quantifying and generating a consensus friction entropy index; a graph feature analysis module for extracting transaction flow graph data and generating risk exposure assessment results; a state machine decision-making module for combining the consensus friction entropy index and the risk exposure assessment results to output a target verification level; and an adaptive verification execution module for executing corresponding verification logic, updating the commodity voucher verification status, and adjusting smart contract resource consumption parameters to provide feedback control over the consensus friction entropy index of the blockchain network, thereby achieving coordinated control of verification efficiency and system operational security.

Description

Technical Field

[0001] This invention relates to the field of blockchain and smart contract application technology, specifically to a smart contract-based anti-counterfeiting information verification system for commodity vouchers. Background Technology

[0002] Existing anti-counterfeiting information verification systems for gift certificates include gift certificate status verification mechanisms based on smart contract execution;

[0003] In related technologies, in order to verify the authenticity of voucher redemption requests, the voucher identifier, account information, signature information and redemption status can usually be verified according to preset rules; or, the historical circulation records of vouchers can be traced and the redemption judgment can be completed by combining the transaction confirmation information in the blockchain network. However, in high-concurrency verification scenarios, the former method is less capable of identifying complex and abnormal transaction behaviors, while the latter method is prone to exacerbating blockchain network congestion due to its long verification chain and high computational overhead. Therefore, the anti-counterfeiting information verification methods for commodity vouchers in related technologies cannot simultaneously achieve verification efficiency, risk identification capability, and network carrying stability. Summary of the Invention

[0004] The purpose of this invention is to provide a smart contract-based anti-counterfeiting information verification system for gift certificates, solving the following technical problems: By integrating network stress perception, graph risk identification, state machine routing, and post-execution feedback adjustment, we can achieve coordinated control over reconciliation efficiency and system data security, and effectively reduce the risk of on-chain congestion caused by over-verification in high-concurrency scenarios.

[0005] The objective of this invention can be achieved through the following technical solutions: A smart contract-based anti-counterfeiting information verification system for gift certificates, deployed in a blockchain node server, includes: The request receiving module is used to receive coupon redemption requests sent by the terminal; The network status monitoring module is used to acquire node computing status data and consensus latency data of the blockchain network, and quantify and generate consensus friction entropy index. The graph feature analysis module is used to extract the transaction flow graph data corresponding to the voucher redemption request, and uses a graph neural network model to parse the transaction flow graph data to generate risk exposure assessment results. The state machine decision module is used to combine the consensus friction entropy index and the risk exposure assessment results to trigger the dynamic routing strategy of the smart contract and output the target verification level. The adaptive reimbursement execution module is used to execute the corresponding reimbursement verification logic based on the target reimbursement verification level, generate execution results, update the reimbursement status of the voucher corresponding to the voucher reimbursement request based on the execution results, and adjust the resource consumption parameters of the smart contract according to the reimbursement status to control the consensus friction entropy index of the blockchain network.

[0006] As a further aspect of the present invention, the method for quantifying and generating consensus friction entropy indicators includes: Extract node computing resource consumption rate from node computing status data; Extract block packaging time difference data from consensus delay data, and calculate the cumulative consensus delay value based on the block packaging time difference data; The node computing resource consumption rate and consensus delay accumulation value are processed to eliminate the dimensions; the preset first weight coefficient and second weight coefficient are obtained, and the node computing resource consumption rate and consensus delay accumulation value after eliminating the dimensions are calculated by weighted summation based on the first weight coefficient and second weight coefficient to generate the consensus friction entropy index. Obtain the historical change sequence of the consensus friction entropy index, use a time series prediction algorithm to calculate the remaining time for the blockchain network to reach the preset consensus friction entropy shutdown threshold based on the historical change sequence, and mark the remaining time as a safe time window.

[0007] As a further aspect of the present invention, the method for extracting transaction flow graph data corresponding to a voucher redemption request includes: Parse the voucher redemption request to obtain the historical transfer records and transaction frequency data of the target voucher corresponding to the voucher redemption request; By using the interactive accounts in the historical transfer records as graph nodes, the transfer behaviors in the historical transfer records as graph edges, and the transaction frequency data as the weight attributes of the graph edges, a local transaction graph is constructed. Extract the topological structure information of local transaction graphs to generate transaction flow graph data.

[0008] As a further aspect of the present invention, the method for analyzing transaction flow graph data using a graph neural network model to generate risk exposure assessment results includes: Input the transaction flow graph data into the graph neural network model; By using the graph convolutional layer in the graph neural network model to perform node feature aggregation and graph pooling operations on the transaction flow graph data, local anomaly features and global topological features of the transaction flow graph data are extracted. Local anomaly features and global topological features are concatenated to generate a comprehensive feature vector; The comprehensive feature vector is input into the pre-defined fully connected classification layer in the graph neural network model to calculate the fraud probability value, and the fraud probability value is mapped to the risk exposure assessment result.

[0009] As a further aspect of the present invention, the method for triggering the dynamic routing strategy of the smart contract and outputting the target verification level includes: Obtain the preset danger threshold and preset time threshold; The consensus friction entropy index is compared with the danger threshold, and the safe time window is compared with the preset time threshold. If the consensus friction entropy index is less than the danger threshold and the safe time window is greater than or equal to the preset time threshold, then the regular verification level is assigned according to the risk exposure assessment results, and the regular verification level is used as the target write-off verification level. If the consensus friction entropy index is greater than or equal to the danger threshold, or the safety time window is less than the preset time threshold, a downgrade routing rule is triggered, and the target verification level is output based on the downgrade routing rule.

[0010] As a further aspect of the present invention, the method for outputting the target verification level based on the downgraded routing rules includes: Obtain a preset controllable risk threshold; Compare the risk exposure assessment results with the controllable risk threshold; If the risk exposure assessment result is less than the controllable risk threshold, a lightweight verification level is generated and used as the target write-off verification level. If the risk exposure assessment result is greater than or equal to the controllable risk threshold, a delay processing level is generated, and the delay processing level is used as the target write-off verification level.

[0011] As a further aspect of the present invention, the method for executing corresponding reconciliation verification logic based on the target reconciliation verification level includes: If the target verification level is lightweight verification level, the basic signature verification logic is triggered, the full-link tracing operation is stopped, and an execution result indicating that the verification has passed is generated. If the target verification level is the delayed processing level, the voucher verification request will be suspended and moved to the preset delayed queue, the current verification operation will be stopped, and an execution result indicating that the request is suspended will be generated. If the target verification level is the standard verification level, the full-chain traceability verification logic is triggered to perform in-depth verification of the issuance and circulation nodes of the voucher and generate the corresponding execution results.

[0012] As a further aspect of the present invention, a method for adjusting the resource consumption parameters of a smart contract based on the write-off status to control the consensus friction entropy index of the blockchain network includes: Obtain the basic execution computing power overhead parameters of the smart contract; The risk exposure assessment results and the consensus friction entropy index are linearly normalized separately and then multiplied to obtain the comprehensive risk value. The comprehensive risk value is mapped to a dynamic penalty coefficient based on a preset positive correlation mapping function; Multiply the basic execution computing power overhead parameter by the dynamic penalty coefficient to obtain the target execution computing power overhead parameter, which serves as the resource consumption parameter; The resource consumption parameters are broadcast to the blockchain network to reduce the number of concurrent requests by increasing the computing power overhead of abnormal transactions, thereby feeding back to control the consensus friction entropy index.

[0013] The beneficial effects of this invention are: 1. This invention triggers a dynamic routing strategy by combining consensus friction entropy index and risk exposure assessment results. When the network is under pressure, requests are automatically diverted to lightweight verification or delayed processing, and full-link tracing is performed when the network is secure. This mechanism effectively avoids on-chain congestion caused by over-verification in high-concurrency scenarios, and improves the rewrite processing speed while ensuring system data security. 2. This invention constructs a local transaction graph that includes interactive accounts and transfer behavior, and uses a graph neural network model to extract local abnormal features and global topological features. Compared with the traditional verification method based on static rules, this solution can accurately capture hidden abnormal circulation patterns such as short-term high-frequency turnover of multiple accounts and cluster forgery, which greatly enhances the anti-counterfeiting and traceability capabilities of the system. 3. This invention quantifies the node computing resource consumption rate and the cumulative consensus delay into a unified consensus friction entropy index, and uses a time series prediction algorithm to calculate the safe time window before system downtime; this mechanism realizes the early prediction of blockchain congestion trends based on preset basic computing power overhead, and upgrades the traditional passive rate limiting to active load reduction based on quantitative time series basis. 4. Based on the risk results associated with the write-off status and the network friction index, this invention dynamically calculates the penalty coefficient and adjusts the resource consumption parameters of the smart contract. This mechanism automatically increases the execution fee for abnormal transactions that are high-risk and exacerbate congestion, and uses the cost of computing power to precisely increase the attack of malicious entities, thereby suppressing abnormal request concurrency at the source and ensuring the long-term stability of the consensus network. Attached Figure Description

[0014] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a schematic diagram of a module of a smart contract-based anti-counterfeiting information verification system for commodity coupons, provided in an embodiment of this application. Detailed Implementation

[0015] 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 some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] Please see Figure 1 A smart contract-based anti-counterfeiting information verification system for commodity vouchers, deployed in a blockchain node server, includes: a request receiving module for receiving commodity voucher verification requests sent by terminals; The network status monitoring module is used to acquire node computing status data and consensus latency data of the blockchain network, and quantify and generate consensus friction entropy index. The graph feature analysis module is used to extract the transaction flow graph data corresponding to the voucher redemption request, and uses a graph neural network model to parse the transaction flow graph data to generate risk exposure assessment results. The state machine decision module is used to combine the consensus friction entropy index and the risk exposure assessment results to trigger the dynamic routing strategy of the smart contract and output the target verification level. The adaptive reimbursement execution module is used to execute the corresponding reimbursement verification logic based on the target reimbursement verification level, generate execution results, update the reimbursement status of the voucher corresponding to the voucher reimbursement request based on the execution results, and adjust the resource consumption parameters of the smart contract according to the reimbursement status to control the consensus friction entropy index of the blockchain network.

[0017] This embodiment provides a smart contract-based anti-counterfeiting information verification mechanism for commodity vouchers. Specifically, the mechanism is deployed in the consortium blockchain environment of the digital consumer voucher distribution platform. It is designed for high-concurrency verification scenarios during high-concurrency events and performs closed-loop control over the entire lifecycle of a single commodity voucher, from submitting a verification request to risk identification, verification level selection, execution of verification, writing back the on-chain state, and reverse adjustment of resource parameters. Specifically, the request receiving module receives the coupon identifier, user account identifier, merchant account identifier, request timestamp, and terminal signature data uploaded by the terminal; For ease of explanation, we can define a single write-off request as a "request". The coupon number is User account is Merchant account is The requested timestamp is ; After receiving the request, the system first performs format validation to confirm that the fields are complete and the signature structure is valid, and then generates a temporary processing identifier corresponding to the request. This is to ensure that subsequent modules can process the same object. The network status monitoring module collects node computing status data and consensus latency data from each accounting node of the consortium blockchain. The node computing status data may include CPU utilization, memory utilization, and the length of the pending transaction queue. The consensus latency data may include proposal generation time, packaging confirmation time, and multi-node confirmation difference. After merging the above data, the system forms a set of network-side features, such as nodes. , , The current computing resource consumption rates are respectively , , The corresponding average value can be denoted as The packaging time differences of the three blocks within the same time window are 1s, 2s, and 3s, respectively, and their cumulative delay can be recorded as 6s. Based on these characteristics, the system generates a consensus friction entropy index that reflects the current network stress level. For use in subsequent decision-making; The graph feature analysis module revolves around the request. Redeem gift certificates Historical data flow; configure interactive accounts as graph nodes, and configure transfer or transaction behaviors as graph edges; if experience in sequence Forward to , Forward to , Initiating a write-off can create a partial transaction graph. ; In this graph, attributes such as transaction frequency, transfer interval, and number of associated devices can be attached to the edges; the system will then display the graph. After inputting a preset graph neural network model, the output is the risk exposure assessment result of the current write-off request for this bond. For example, the model outputs a fraud probability of ; After segmentation and mapping, a medium-to-high risk assessment result can be obtained; The state machine decision module receives the consensus friction entropy index from the network side. Risk exposure assessment results on the business side It performs dynamic routing based on a preset state machine; the state machine can maintain at least three states: regular verification state, lightweight verification state, and delayed processing state. For ease of explanation, if At a low level and If the value is low, it flows into the regular verification state; if High but If it remains within a controllable range, switch to lightweight verification mode; if It is close to the cybersecurity boundary and If the level is also high, switch to delayed processing mode; the output result is the target verification level. ; The adaptive reconciliation execution module determines the reconciliation level based on the verification level. Execute the corresponding reconciliation logic; when When the verification level is normal, the system triggers full-chain traceability verification to deeply verify the consistency between the issuing account, distribution account, historical transfer account, and current submitting account of the voucher; when At the lightweight verification level, only basic signature verification, voucher validity verification, and no duplicate redemption verification are performed; when When the processing level is delayed, the request is placed in a delay queue and processing will resume after the network load decreases. After execution, the system generates execution results, such as "redeem approved," "redeem rejected," or "request suspended," and updates the coupon status synchronously. Change from "eligible for cancellation" to "cancelled" or "pending review"; After the state is updated, the system does not end the process, but adjusts the resource consumption parameters of the smart contract in reverse according to the execution result and the current risk level. For example, for similar requests that trigger abnormal graphs more than a preset frequency threshold within a preset time window, increase the execution computing power overhead parameters or add computing resource requirements to suppress abnormal concurrency; for similar requests that trigger abnormal graphs less than or equal to the preset frequency threshold within a preset time window, keep their corresponding resource consumption parameters unchanged; keep the basic resource parameters unchanged for requests that pass normally and have low risk, so as to avoid indiscriminately increasing the node processing occupancy rate of ordinary requests. Once this parameter is broadcast to the consortium blockchain network, it will affect the speed and execution priority of subsequent transactions entering the pool, thereby influencing the consensus friction entropy index. Feedback control closed loop; As an exception handling mechanism, if the terminal signature is missing, the field is incomplete, or the timestamp exceeds the allowed drift range during the request receiving stage, a rejection result is generated directly and the exception log is recorded. If the network status monitoring module does not obtain enough node data within a certain collection period, the monitoring result of the most recent valid window is used and a low confidence mark is attached to prevent the system from losing decision input due to a single collection failure. If the graph feature analysis module cannot form an effective graph structure due to the historical flow records being lower than the preset threshold for the number of nodes, it degenerates into a risk assessment process based on basic rules. If the state machine undergoes a state transition at the threshold boundaries of two levels, a minimum hold time can be introduced to maintain the previous state within a set number of seconds to avoid frequent switching. Within the pre-set peak verification window, merchants The store simultaneously received tens of thousands of digital voucher redemption requests, among which... Corresponding product coupons It has been transferred twice in the last 10 minutes, and the graph model has identified that it has medium to high risk characteristics. At the same time, the consortium blockchain has detected that the average computing resource consumption rate of nodes has risen to [a certain level]. As block confirmation delays continue to rise, the system no longer performs deep tracing in all cases. Instead, the state machine determines the processing path based on the network stress level and the risk level of the bond. If it is determined that there is still a safety margin, then... Perform routine verification; if the system is about to approach the congestion boundary, reduce the risk of requests to lightweight verification and transfer high-risk requests to the delay queue; in this way, each voucher under the same business line can be included in a unified and continuous closed-loop processing. The purpose of this mechanism is to integrate network stress perception, graph risk identification, state machine routing, and post-execution feedback adjustment into a single system in the context of anti-counterfeiting verification of commodity vouchers. This will enable coordinated control of verification efficiency and system data security, and reduce the risk of on-chain congestion caused by over-verification in high-concurrency scenarios.

[0018] In a preferred embodiment of the present invention, the method for quantifying the consensus friction entropy index includes: extracting the node computing resource consumption rate from node computing state data; extracting block packaging time difference data from consensus delay data; and calculating the cumulative consensus delay value based on the block packaging time difference data. The node computing resource consumption rate and consensus delay accumulation value are processed to eliminate the dimensions; the preset first weight coefficient and second weight coefficient are obtained, and the node computing resource consumption rate and consensus delay accumulation value after eliminating the dimensions are calculated by weighted summation based on the first weight coefficient and second weight coefficient to generate the consensus friction entropy index. Obtain the historical change sequence of the consensus friction entropy index, use a time series prediction algorithm to calculate the remaining time for the blockchain network to reach the preset consensus friction entropy shutdown threshold based on the historical change sequence, and mark the remaining time as a safe time window.

[0019] This embodiment provides a mechanism for quantitatively characterizing network operating pressure. Specifically, in the above-mentioned write-off mainline, if only the node computing resource occupancy rate is higher than a preset threshold or the consensus delay is greater than a preset delay threshold for qualitative judgment, it is difficult to support the subsequent state machine to make stable routing decisions. Therefore, it is necessary to compress multi-source network state data into a single, comparable, and predictable consensus friction entropy index, and further calculate the safe time window of the system from the shutdown boundary. Specifically, the system collects data from the computation and consensus sides of the consortium blockchain nodes at a fixed sampling period; on the computation side, the node's computational resource consumption rate can be selected as the core quantity, for example, for a node... , , The current CPU utilization rates are 72%, 84%, and 78%, respectively, which average to 0.78. If memory usage and queue length still need to be included, they can be converted into a uniform consumption rate according to a preset ratio before being incorporated into the calculation; the consensus side extracts the packaging time difference data of multiple blocks within the same time window. Assuming the theoretical standard packet interval is 2s, while the actual confirmation time for three consecutive blocks is 3s, 4s, and 5s, the corresponding time differences are 1s, 2s, and 3s, and the cumulative consensus delay can be recorded as 6s. After obtaining the node computing resource consumption rate and the cumulative consensus delay, the system introduces a first weight coefficient and a second weight coefficient to form a weighted sum of the two types of factors. For ease of explanation, the first weighting coefficient can be set to The second weighting coefficient is set to Normalize the cumulative consensus delay value to a preset upper limit of 10 seconds, then the normalized value after 6 seconds is: At this point, the consensus friction entropy index Press Calculations were performed to obtain ; The higher this value, the closer the network is to the congestion boundary caused by complex reconciliation logic; obtaining only the current value is insufficient to support early intervention, therefore the system also obtains the historical change sequence of the consensus friction entropy index; for example, in the most recent 5 sampling windows, In order , , , , ; The system uses time series forecasting algorithms to predict its continuous trend; lightweight methods such as moving average, exponential smoothing, or first-order regression can be used here. Taking the first-order regression algorithm as an example, the system uses the sampling period number in the historical change sequence as the independent variable and the corresponding consensus friction entropy index as the dependent variable. The slope of the straight line is calculated by fitting the least squares method. This slope represents the rate of increase of the entropy value in each sampling period. Assuming the currently calculated upward slope is approximately The preset shutdown threshold is The system subtracts the latest indicator from the preset shutdown threshold. The remaining entropy space is obtained. ; Divide the remaining entropy space by the slope value This allows for the accurate calculation of approximately 4 sampling cycles remaining before the shutdown boundary is reached; if each sampling cycle is 30 seconds, the safe time window is approximately 120 seconds. In implementation, the safe time window can be written into the state machine's shared buffer for subsequent verification level selection; thus, the system no longer relies on the passive mode of degrading only when it is already blocked, but instead turns into an active mode of releasing pressure in advance when it is predicted to be blocked. As an anomaly tolerance mechanism, if there are missing node reports in a certain sampling period, the system will calculate based on valid samples with no less than the preset minimum number of nodes; if there are insufficient valid samples, the index results of the previous period will be used and an incomplete data mark will be marked internally, prompting the state machine to make a more conservative decision. If the upper limit of the denominator is configured to 0 or is abnormally missing when normalizing the consensus delay accumulation value, it will automatically fall back to the factory default upper limit to prevent calculation interruption. If the length of the historical sequence is insufficient to support the prediction, for example, if only two windows are sampled when the system is first started, the predicted value of the safe time window will not be output for the time being, but the unknown window state will be output instead. In this case, the state machine can be processed by default according to the neutral to conservative strategy. Eight minutes after the start of peak business hours, the platform detected a decrease in the redemption speed for regular users, but no widespread failures had yet occurred. At this time, the off-chain monitoring service calculated the average computing resource consumption rate of the nodes. The consensus delay cumulative value after normalization is The consensus friction entropy index was calculated to be: ; Combining the ascending trajectories of multiple preceding windows, the system predicts the distance. The platform has only 120 seconds of safe time left before the shutdown boundary is reached. Based on this result, the platform can limit the entry speed of complex verification requests in advance before a real transaction backlog occurs, thus ensuring the computing resources for the subsequent anti-counterfeiting verification mechanism. The purpose of this step is to establish a unified measure of the stress state of the blockchain network with low computational overhead, and to provide quantifiable timing basis for subsequent dynamic routing and load reduction strategies through the prediction of safe time windows.

[0020] In a preferred embodiment of the present invention, the method for extracting transaction flow graph data corresponding to a voucher redemption request includes: parsing the voucher redemption request and obtaining historical transfer records and transaction frequency data of the target voucher corresponding to the voucher redemption request; By using the interactive accounts in the historical transfer records as graph nodes, the transfer behaviors in the historical transfer records as graph edges, and the transaction frequency data as the weight attributes of the graph edges, a local transaction graph is constructed; the topological structure information of the local transaction graph is extracted to generate transaction flow graph data.

[0021] This embodiment provides a transaction flow graph construction step; specifically, in the above main line, if only the voucher code and account in a single reimbursement request are used for verification, although direct anomalies such as duplicate reimbursement can be identified, it is difficult to discover continuous transfers of multiple accounts and multiple nodes and high-frequency transactions within a preset time window. Therefore, it is necessary to convert the historical flow relationship of commodity vouchers into graph structure data that can be analyzed by the model. Specifically, after receiving a redemption request, the system parses the unique identifier of the target voucher and traces the voucher's historical transfer records and transaction frequency data in the business database and on-chain transaction records; Assuming a gift certificate Three key phases are formed during its life cycle: Forward to , Forward to , Forward to ,and and If there are 5 high-frequency interactions of the same type of coupon within the last hour, the system will organize this information into a raw flow set associated with the current request. The system constructs a local transaction graph; here, , , , Each is used as a node in the graph; As a directed edge; Transaction frequency data is used as the weight attribute of the edges; for ease of representation, a local topology structure containing 4 nodes and 3 edges is constructed; if further represented by an adjacency matrix, the first row and third column, which are not directly connected, can be recorded as 0, and the second row and third column, due to... arrive The existence of a transfer relationship can be recorded as 1, and the edge weight is 5, indicating that this type of interaction is more frequent; if a time interval is also required on the edge, the edge attribute can be expanded into a binary feature with weight = 5 and the most recent interval = 10 minutes. After the local transaction graph is generated, the system extracts its topological structure information to form transaction flow graph data. The extracted content may include the number of nodes, the number of edges, the average out-degree, the maximum in-degree, whether there is backflow, whether there is short-cycle closed loop, path depth, etc. To simplify the example, if The circulation chain depth is 3 and the maximum out-degree is 1. However, there are multiple vouchers that cross and travel between related accounts, which can be further marked as suspected cluster structures. The system encapsulates these structured results together with attributes such as edge weights and timestamps as input data for subsequent graph neural network models. Compared to a single transaction record, this graphical representation can preserve the direction and density information of the relationships between accounts, which helps to improve the dynamic analysis from static verification of whether the voucher itself is valid to whether the voucher's circulation method is abnormal. As an anomaly tolerance mechanism, if the target commodity voucher only has two records, issuance and current holding, and cannot form a complex graph structure, the system will still construct a minimal local transaction graph, whose nodes include at least the issuing account and the current redemption account, and fill the missing parts with default edge attributes; if there is duplicate on-chain data in the historical record for the same period, the graph will be constructed after deduplication based on the transaction hash. If certain account identifiers cannot be directly displayed due to privacy anonymization, the anonymized unified mapping identifier can be used as a node without affecting the structural analysis. If a break in the flow record is found during the extraction process, such as a missing confirmation block in the middle transfer, the edge is marked as a low-confidence edge so that the subsequent model can reduce its contribution. During high-concurrency processing, malicious attack nodes often use multiple ordinary accounts to disrupt the circulation of a single coupon in order to circumvent traditional rules; for commodity coupons On the surface, only three normal transfers occurred, but after mapping, it can be found that... and Within a short period, multiple similar coupons were frequently transferred between each other, resulting in an unusually dense cluster structure in some transaction patterns; based on this, the system no longer records them. Instead of treating it as a regular single-chain transfer, it is included in complex transfer objects to provide a more discernible data foundation for subsequent risk assessment; The purpose of this step is to transform the historical circulation relationship of coupons from discrete records into graph structure data with direction, frequency and topological semantics, so as to realize the computable expression of hidden abnormal circulation patterns.

[0022] In a preferred embodiment of the present invention, the method for analyzing transaction flow graph data using a graph neural network model to generate risk exposure assessment results includes: inputting the transaction flow graph data into the graph neural network model; By using the graph convolutional layer in the graph neural network model to perform node feature aggregation and graph pooling operations on the transaction flow graph data, local anomaly features and global topological features of the transaction flow graph data are extracted. Local anomaly features and global topological features are concatenated to generate a comprehensive feature vector. The comprehensive feature vector is then input into a pre-defined fully connected classification layer in the graph neural network model to calculate the fraud probability value, which is then mapped to the risk exposure assessment result.

[0023] This embodiment provides a risk exposure assessment step based on graph neural networks. Specifically, after the aforementioned graph construction is completed, if manually set static rules are continued, such as high risk for more than 3 transfers, it is easy for organized accounts to circumvent these rules by dispersing paths and controlling frequency. Therefore, it is necessary to introduce a graph neural network model that can simultaneously understand local connection patterns and overall structural forms to automatically parse the transaction flow graph data. Specifically, the complete structure of the graph neural network model includes sequentially connected graph convolutional layers, graph pooling layers, feature concatenation layers, and fully connected classification layers. The graph convolutional layers aggregate node features based on the topological structure of the local transaction graph, and their single-layer graph convolutional feature propagation formula is: In the formula, Representing the The node feature matrix of the layer, This represents the introduction of a self-connected local transaction graph adjacency matrix. This represents the degree matrix corresponding to the adjacency matrix. The weight matrix represents the learnable weight matrix in the graph convolutional layer. Represents a non-linear activation function; Specifically, the system first standardizes the local transaction graph into an input format that the model can recognize; for each node, node characteristics can be set as account type, account age, number of write-offs in the past 7 days, etc. For each edge, edge characteristics can be set such as transfer frequency, time interval, and device overlap; using a graph. For example, if node The feature vector is simplified to [short account age, high activity in the last 7 days], node If the feature vector is simplified to [high device overlap, many cross-merchant redemptions], then this information can be used as the initial node representation input model; Graph convolutional layers aggregate features from adjacent nodes; A very simplified deduction can be made: If The initial value of the inherent risk characteristics is Its upstream node for downstream nodes for After one layer of aggregation, The feature representation can be generated based on a weighted average operation, i.e. ; By aggregating features from neighboring nodes, the model can extract graph nodes whose local features are normal but whose topological neighborhoods have abnormal features. Further, through graph pooling, the system compresses the entire local graph into a global representation. When extracting global topological features, the system performs full-graph-dimensional data compression and feature extraction on all node feature vectors after aggregation through operations such as max pooling or mean pooling. For example, by using a max pooling strategy, the maximum value is extracted from the risk-related feature dimension of all nodes, so as to determine whether there are cluster nodes with a topological density greater than a preset threshold in the entire local graph. Simultaneously, mean pooling is used to reflect the overall connectivity activity of the region. After the pooling operation, a 4-node small graph can be stably compressed into a fixed-length global topology vector to describe the overall characteristics such as graph depth, cluster density, and path concentration. In terms of extraction results, the model can output both local anomaly features and global topological features. Local anomaly features focus more on the abnormal interactions between individual nodes and key edges, while global topological features focus more on whether the entire graph exhibits patterns such as clustering, chain transfers, or star-shaped diffusion. The system concatenates the two into a comprehensive feature vector, for example, a local anomaly feature vector is... The global topological feature vector is After concatenation, a comprehensive feature vector is generated. The integrated feature vector is input into a pre-defined fully connected classification layer, and the output is a fraud probability value. ; For ease of explanation, if the classification layer outputs... Then, according to the preset mapping relationship, 0 to 0.3 can be mapped as low risk, 0.3 to 0.6 as medium risk, and above 0.6 as high risk. Mapped to risk exposure assessment results characterizing high risk; To adapt to subsequent modules, this risk exposure assessment result can be further converted into a numerical form that can be directly invoked by the state machine, for example, denoted as... ; In terms of model deployment location, it is preferable to place it off-chain analysis nodes for execution, and the on-chain only receives its result summary or signature result, thereby avoiding the direct inclusion of complex graph computation into the consensus process computation; As an exception handling mechanism, if the number of nodes in the input graph is lower than the preset node number threshold, causing deep graph convolution to easily lose its meaning, the system can reduce the number of convolutional layers, retaining only one aggregation layer and one classification layer. If certain node features are missing, such as account age not being successfully synchronized, the default value will be filled in and a missing flag will be added to prevent the entire image from being discarded directly. If the fraud probability value output by the model is exactly at the mapping boundary, for example, exactly 0.60, it can be classified into a higher level of risk according to the preset strategy to avoid underestimating the risk under the critical value; if the model service is temporarily unavailable, it will revert to the emergency assessment method based on the graph statistical rules, but the unified risk exposure assessment result output format will still be retained. During peak business periods and peak redemption times, gift vouchers The partial map did not show obvious signs of forgery, but its upstream accounts and transit accounts It forms short-term, intensive interactions with multiple newly registered accounts; the graph convolutional layer aggregates these neighborhood features and identifies... Although each individual action appears normal, it is deeply embedded in the core of an abnormal cluster; Graph pooling also captured the unnatural clustering and compression structure in the entire image; the final classification layer output... Based on the probability of fraud, the system marks the risk exposure assessment result of the current redemption request of the voucher as high risk, providing a basis for whether to perform in-depth verification in the future. The purpose of this step is to use graph structure learning capabilities to extract hidden risk patterns from complex transaction flow relationships, thereby achieving a more adaptive risk assessment than static rules.

[0024] In a preferred embodiment of the present invention, the method for triggering the dynamic routing strategy of the smart contract and outputting the target verification level includes: obtaining a preset danger threshold and a preset time threshold; comparing the consensus friction entropy index with the danger threshold and comparing the safe time window with the preset time threshold. If the consensus friction entropy index is less than the danger threshold and the safe time window is greater than or equal to the preset time threshold, then the regular verification level is assigned according to the risk exposure assessment results, and the regular verification level is used as the target write-off verification level. If the consensus friction entropy index is greater than or equal to the danger threshold, or the safety time window is less than the preset time threshold, a downgrade routing rule is triggered, and the target verification level is output based on the downgrade routing rule.

[0025] This embodiment provides a dynamic routing step oriented towards the system's survival boundary; specifically, in the aforementioned mechanism, if the system decides whether to perform deep verification based solely on the risk exposure assessment results, it is easy to push a large number of complex requests onto the chain for execution during network congestion, which amplifies the consensus pressure. Therefore, in addition to risk assessment, this embodiment introduces two network-side constraints: consensus friction entropy index and security time window, to achieve a routing strategy that first determines whether the system is capable of performing full-link source tracing verification, and then decides whether to perform full-link source tracing verification. Specifically, the system pre-configures danger thresholds. With time threshold For ease of explanation, the danger threshold can be set to 0.75, and the time threshold can be set to 180s; the consensus friction entropy index corresponding to the current request received by the state machine. With safe time window Then, perform two levels of judgment; The first step is to determine whether the system is in a safe carrying zone; if and s indicates that the network still has sufficient processing capacity, and at this time the state machine is allowed to enter the regular verification path according to the risk exposure assessment results; For example, for risk exposure The system can still perform deep tracing even with a request, because the current network can still bear the additional verification overhead; at this time, the output target verification level is the normal verification level. The second step is to determine whether the system has entered the pressure zone; as long as it appears... or Any of the conditions in s indicates that the network is no longer suitable for continuing to perform complex verification on a large scale; At this point, the state machine no longer directly issues the regular verification level, but instead triggers the downgrade routing rule, handing over subsequent judgments to a more conservative process; and enters a system blocking state that exceeds the consensus friction entropy shutdown threshold. For ease of understanding, a simplified deduction can be made; if at a certain moment , If s, then the safe carrying area conditions are met, and the system retains its regular verification capabilities; If at another moment , s, although the remaining time has not yet fallen below the threshold, the system immediately enters a degraded routing state because the consensus friction entropy has exceeded the danger threshold; for example, , Even though the current entropy value has not yet exceeded the limit, the prediction indicates that the network is about to enter the danger zone, thus triggering a degraded route as well; this dual-condition judgment can avoid the lag of a single indicator. As an anomaly tolerance mechanism, if a danger threshold or time threshold is missing due to a configuration error, the system will use the factory default value and record the configuration anomaly; if and If both are unavailable, the state machine defaults to conservative mode and does not allow normal verification directly; if Exactly equal to the danger threshold, or If the time threshold is exactly equal to the threshold, then it is handled as either reaching the danger boundary or still meeting the time boundary, to ensure consistency of the boundary rules. If different monitoring nodes report If there is a significant disagreement on the value, the shortest time window should be taken as the basis for decision-making to avoid overestimating the safety margin. At the 10-minute mark of peak business activity, the platform detected that the consensus friction entropy index had risen to [a certain level]. Although the trend predicts that there are still 210 seconds before the shutdown boundary, the network is already under obvious pressure. At this time, for gift certificates For requests with complex graphs, the system no longer maintains the original deep verification strategy, but instead uniformly triggers the degradation routing rule; conversely, if it is in the early stages of the activity... Only and If the time limit is 600 seconds, even if some requests are of medium to high risk, the system can still use relatively sufficient resources to perform a complete source tracing. The purpose of this step is to incorporate network capacity into the verification level selection process, thereby enabling synchronized adjustment of the revocation path and the real-time stress status of the system, and reducing on-chain overload caused by fixed verification strategies.

[0026] In a preferred embodiment of the present invention, the method for outputting the target verification level based on the degradation routing rule includes: obtaining a preset controllable risk threshold; and comparing the risk exposure assessment result with the controllable risk threshold. If the risk exposure assessment result is less than the controllable risk threshold, a lightweight verification level is generated and used as the target write-off verification level. If the risk exposure assessment result is greater than or equal to the controllable risk threshold, a delay processing level is generated, and the delay processing level is used as the target write-off verification level.

[0027] This embodiment provides a detailed decision-making mechanism within the degraded routing; specifically, in the previous embodiment, the system was already able to identify the situation where the current network is not suitable for continuing to fully perform full-link source tracing verification, but if all requests are suspended once they enter the degraded routing, the success rate of normal business processing will be lower than the preset threshold. If all requests are downgraded to the lightweight verification level, the probability of high-risk requests bypassing the anti-counterfeiting verification node may increase, thereby increasing the risk of system operation. Therefore, this embodiment further introduces a controllable risk threshold to redistribute requests in the downgraded state. Specifically, the system pre-sets controllable risk thresholds. This is used to distinguish between two categories of objects: those where the risk exists but is acceptable at the current stage, and those where the risk has exceeded the current acceptable range. For ease of explanation, it can be said that Set as After entering the degraded route, the state machine reads the risk exposure assessment result of the write-off request. and with Compare; when This indicates that although the request may not be entirely low-risk, its risk level is still within the system's controllable range; At this point, a lightweight verification level is generated; for example... In response to the request, the system can temporarily waive the full-link tracing and retain only the necessary verification in order to optimize the node processing speed and release the system's concurrent computing resources; when If the request is mistakenly allowed during a period of resource scarcity, it could pose a system operational risk exceeding the capacity of the safe time window. In this case, the system generates a delayed processing level, moves the request out of the real-time main channel, and waits for the network to recover before conducting a more rigorous review. For example gift certificates During periods of high consensus friction entropy, it is not advisable to directly allow passage through lightweight paths; instead, priority should be given to protecting the security of the system and its data. Further logical deduction; assuming there are three requests entering the degraded route within a certain minute, the risk exposure assessment results are as follows: , , ;exist Under the conditions, and It was assigned to the lightweight verification level. They are assigned to the delayed processing level; in this way, the system will neither cause the reconciliation channel to stop abruptly due to a complete suspension, nor expose excessive risks due to a complete simplification of verification. As an anomaly tolerance mechanism, if the risk exposure assessment results are missing, such as when the map model is temporarily unavailable, the system can use a conservative compensation value for comparison, preferably considering it as not lower than the controllable risk threshold, to prevent high-risk requests from being mistakenly allowed. like Insufficient configuration can cause a large number of requests to enter the delay queue. The system can trigger threshold calibration based on the queue backlog, but this calibration should ideally be subject to manual approval or automatic adjustment within a limited scope to prevent reverse manipulation during attacks. Exactly equal to If the risk level is high, it will be processed according to the high-risk side, and the delayed processing level will be directly output to reduce boundary ambiguity. During peak periods of high-concurrency business, the platform has entered a degraded routing state; for a batch of pending redemption requests, one of the target redemption requests corresponds to a voucher that has only undergone one normal transfer, and its risk exposure assessment result is 45. The system sends it to the lightweight verification path so that ordinary users can still complete the basic redemption that meets the preset delay requirements. And gift coupons Because it involves a cluster of accounts, the risk exposure assessment result is as follows: The system then moves it to a delay queue; in this way, at the same time, the platform can prioritize the use of limited computing resources for global stability, while avoiding a uniform suspension strategy that applies to all revocation requests indiscriminately. The purpose of this step is to further subdivide the risk levels under network pressure, thereby achieving a balance between real-time reconciliation efficiency and acceptable risk range.

[0028] In a preferred embodiment of the present invention, the method of executing the corresponding verification logic based on the target verification verification level includes: if the target verification verification level is a lightweight verification level, then triggering the basic signature verification logic, stopping the execution of the full-link tracing operation, and generating an execution result indicating that the verification has passed; If the target verification level is the delayed processing level, the voucher verification request will be suspended and moved to the preset delayed queue, the current verification operation will be stopped, and an execution result indicating that the request is suspended will be generated. If the target verification level is the standard verification level, the full-chain traceability verification logic is triggered to perform in-depth verification of the issuance and circulation nodes of the voucher and generate the corresponding execution results.

[0029] This embodiment provides a verification execution step that corresponds one-to-one with the verification level. Specifically, after the aforementioned state machine has output the target verification level, if the execution layer still adopts a unified verification process, the front-end traffic splitting will lose its meaning. Therefore, it is necessary to map different levels to differentiated execution logic so that each level corresponds to clear resource overhead and processing consequences. Specifically, when the target verification level is lightweight verification, the system triggers basic signature verification logic; this logic includes at least the following basic items: terminal signature validity verification, whether the voucher status is unverified, whether the voucher is within its validity period, and whether the submitting account is consistent with the current voucher-holding account. For ease of explanation, regarding the request If the signature verification passes, the coupon code status is normal, the validity period has not expired, and the account is consistent, the redemption execution result will be generated directly, and the coupon status will be changed to redeemed. Unlike conventional verification, this path no longer traces all historical flow nodes, thus significantly reducing the computational burden of joint on-chain and off-chain operations. When the target verification level is the delayed processing level, the system does not perform real-time verification, but suspends the request and writes it to a preset delayed queue; this queue can be organized by fields such as entry time, risk exposure level, and merchant priority. The system generates a request suspension execution result for the request and returns the corresponding suspension flag for subsequent querying by the merchant and user. Suspension does not mean rejection, but rather waiting for the network pressure to subside or for the conditions for manual review to be met before re-entering the processing flow. When the target verification level is the regular verification level, the system triggers the full-link traceability verification logic. In addition to the basic items of lightweight verification, this logic also verifies whether the source of the voucher is an authorized contract, whether the distribution path is within the permitted scope, whether there are abnormal clusters in the historical circulation nodes, and whether there is any cross-correlation with blacklisted accounts. If all deep checks pass, the write-off is approved; if a broken issuance link, duplicate transfer conflict, or abnormal cluster association is found, the write-off is rejected. To more intuitively illustrate the differences between the three types of logic, a logical deduction is performed; if the coupon... If the risk is low and the network is in a secure zone, then perform the standard verification, checking a total of 10 rules. If the gift certificate If the risk is moderate but the network is under pressure, then lightweight verification will be performed, checking only the four basic rules; if it is a voucher... If a system is both at high risk and experiencing congestion, it should be placed in a delay queue to avoid consuming more consensus resources in the current window. As an exception handling mechanism, if the basic signature is found to be invalid during the lightweight verification stage, even if it was originally allocated to the fast channel, a verification rejection should be immediately output instead of allowing it to proceed. If a delayed queue write fails, for example due to a queue service exception, the system can roll back the request to the local persistent cache and mark it as pending write to prevent request loss. If some historical node data is temporarily unavailable during routine verification, the system may temporarily delay issuing a pass result and preferably change the judgment to suspension in order to avoid mistakenly releasing the case when there is insufficient evidence. If a new redemption request is received for the same voucher during the pending period, the subsequent request should be directly identified as a conflicting request and directed to the same pending record to prevent concurrent duplicate queuing. During the nighttime redemption window at peak business hours, a supermarket store submitted three types of requests in succession: The first type was low-risk coupons that ordinary users redeemed immediately after making on-site purchases, which the system quickly approved after only performing signature and validity period verification; The second category is merchandise vouchers. For high-risk vouchers involving anomalous clusters, the system generates a pending result and moves it to a delayed queue. The third category is complex vouchers that are reprocessed after network recovery. The system re-executes full-link tracing and decides whether to release the vouchers based on the results. As a result, different vouchers in the same business line can enter different verification depths based on network status and risk status. The purpose of this step is to ensure that the verification level is truly implemented as an executable and distinguishable write-off logic, thereby achieving a match between resource investment and risk level control.

[0030] In a preferred embodiment of the present invention, the method for adjusting the resource consumption parameters of a smart contract according to the write-off status to control the consensus friction entropy index of the blockchain network includes: obtaining the basic execution computing power overhead parameters of the smart contract; performing linear normalization on the risk exposure assessment result and the consensus friction entropy index respectively, and then multiplying them to obtain a comprehensive risk value; and mapping the comprehensive risk value to a dynamic penalty coefficient based on a preset positive correlation mapping function. The target execution computing power overhead parameter is obtained by multiplying the basic execution computing power overhead parameter by the dynamic penalty coefficient. The resource consumption parameter is then broadcast to the blockchain network to reduce the number of concurrent requests by increasing the computing power overhead of abnormal transactions, and to provide feedback control to the consensus friction entropy index.

[0031] This embodiment provides a feedback control mechanism for dynamically adjusting the resource consumption parameters of smart contracts. Specifically, in the aforementioned embodiments, the system is already able to identify and route single reimbursement requests. However, if the system only passively diverts requests at the execution layer without adjusting the entry cost of subsequent requests, the abnormal entity can continue to generate complex transaction requests, continuously increasing the consensus friction entropy. Therefore, this embodiment dynamically adjusts the execution computing power overhead parameters to impose higher computing resource overhead on requests during high-risk and high-congestion phases, thereby creating a feedforward suppression effect; Specifically, the redemption status not only reflects the current business processing result, but is also a key anchor point that triggers the adjustment of resource consumption parameters. After the system updates the redemption status of the coupon, such as updating it to redeemed, pending review, or suspended, it listens for and captures the status change event. Based on the execution context corresponding to this write-off status, the system extracts the current target processing environment parameters that triggered this status, thus locking in the original risk exposure assessment results for this request. With the latest consensus friction entropy index Based on this, the system first obtains the basic execution computing power overhead parameters of the smart contract. ; This parameter represents the default computing power cost required to perform a standard reconciliation operation in a consortium blockchain; for ease of explanation, it can be set to... Each resource consumption quantification unit; The system assesses the risk exposure results separately. Friction Entropy Index Perform linear normalization; assume The original range is to ,but After normalization ; The original range is to ,but Can be directly regarded as ; The system multiplies the two values ​​to obtain the overall risk value. The system uses a preset positive correlation mapping function to calculate the comprehensive risk value. Mapped to dynamic penalty coefficient ; The mapping function can be a linear function or a piecewise function; for example, in a simplified linear manner, we can assume... ,but If a segmented approach is adopted, then when hour ,when hour ,when hour Regardless of the form used, it should be maintained The larger, The higher; The system will use basic execution computing power overhead parameters With dynamic penalty coefficient Multiply to obtain the target execution computing power overhead parameter. ; Taking the aforementioned linear approach as an example, The resource consumption quantification unit can be rounded down to 15 or 16 according to the on-chain billing precision; the target execution computing power overhead parameter is used as the resource consumption parameter for the processing computing power overhead control of subsequent similar requests; During the broadcast phase, the system can synchronize new resource consumption parameters to the relevant nodes or contract execution environment of the consortium blockchain. After the broadcast, for nodes that submit complex transfer requests frequently in a short period of time, the system increases the computing power cost of their corresponding write-off processing, thereby reducing the probability that such nodes will continuously trigger high computational complexity verification logic. For low-risk, low-congestion scenarios, requests will still maintain a cost close to basic resource consumption to avoid increasing the additional resource burden on ordinary users. This can also be combined with the revocation status to form a closed loop; for example, if most of a batch of requests are suspended or rejected after processing, and their common feature is high-risk cluster accounts, the system can continue to maintain a high penalty coefficient for a period of time; conversely, if the network recovers, the pass rate stabilizes and the consensus friction entropy decreases, the dynamic penalty coefficient can gradually fall back, restoring the normal resource consumption cost. As an anomaly tolerance mechanism, if an anomaly is found during normalization... or If the value exceeds the preset range, it will be truncated to the boundary value before participating in the calculation. For example... according to deal with, according to Processing; if the basic execution computing power overhead parameters If the value is missing, a default value will be used temporarily to prevent the resource consumption calculation rules from being interrupted. If the mapping function is configured incorrectly, it will cause... Less than Then it will be automatically promoted to This ensures that the computational overhead in high-risk scenarios is not reduced in reverse; if some nodes fail to update in time during the broadcast process, version number verification can be attached during contract execution, and a compatible computing power deduction mechanism or refusal to execute can be adopted for nodes with old versions to avoid parameter inconsistencies in the network. During peak periods of high-concurrency business, vouchers A series of unusually complex write-off requests continued to pour in. At this point, the system updated the status of the request to be suspended and locked the risk exposure assessment result associated with the request to 68, and the consensus friction entropy index to 0.78. After normalization and product operations, a comprehensive risk value of 0.5304 is obtained, which maps to a dynamic penalty coefficient of more than 1.5, increasing the target execution computing power overhead parameter of this type of request from the base value of 10 to about 15 resource consumption quantization units. As the resource consumption costs caused by abnormal transactions increase, the number of accounts repeatedly initiating complex write-offs in a short period of time decreases, and network congestion begins to ease; after the consensus friction entropy decreases at the end of the activity, the system will gradually restore the execution computing power overhead parameters to the basic level. The purpose of this mechanism is to transform risk identification results and network congestion status into executable resource allocation threshold adjustment signals, thereby suppressing abnormal concurrent requests in advance and controlling the upward trend of consensus friction entropy through cost feedback.

[0032] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A smart contract-based anti-counterfeiting information verification system for commodity vouchers, characterized in that, Deployed in the blockchain node server, including: The request receiving module is used to receive coupon redemption requests sent by the terminal; The network status monitoring module is used to acquire node computing status data and consensus latency data of the blockchain network, and quantify and generate consensus friction entropy index. The graph feature analysis module is used to extract the transaction flow graph data corresponding to the commodity voucher redemption request, and to use a graph neural network model to parse the transaction flow graph data to generate risk exposure assessment results. The state machine decision module is used to combine the consensus friction entropy index with the risk exposure assessment result to trigger the dynamic routing strategy of the smart contract and output the target verification level. The adaptive reimbursement execution module is used to execute the corresponding reimbursement verification logic based on the target reimbursement verification level, generate an execution result, update the reimbursement status of the voucher corresponding to the voucher reimbursement request based on the execution result, and adjust the resource consumption parameters of the smart contract according to the reimbursement status to provide feedback control over the consensus friction entropy index of the blockchain network.

2. The smart contract-based anti-counterfeiting information verification system for commodity coupons according to claim 1, characterized in that, Methods for quantifying consensus friction entropy metrics include: Extract the node computing resource consumption rate from the node computing status data; Extract block packaging time difference data from the consensus delay data, and calculate the cumulative consensus delay value based on the block packaging time difference data; The node computing resource consumption rate and the consensus delay accumulation value are subjected to dimensionless processing; a preset first weight coefficient and a second weight coefficient are obtained; based on the first weight coefficient and the second weight coefficient, the dimensionless node computing resource consumption rate and the consensus delay accumulation value are weighted and summed to generate the consensus friction entropy index. Obtain the historical change sequence of the consensus friction entropy index, use a time series prediction algorithm to calculate the remaining time for the blockchain network to reach the preset consensus friction entropy shutdown threshold based on the historical change sequence, and mark the remaining time as a safe time window.

3. The smart contract-based anti-counterfeiting information verification system for commodity vouchers according to claim 1, characterized in that, The method for extracting the transaction flow graph data corresponding to the voucher redemption request includes: Parse the voucher redemption request to obtain the historical transfer records and transaction frequency data of the target voucher corresponding to the voucher redemption request; Using the interactive accounts in the historical transfer records as graph nodes, the transfer behaviors in the historical transfer records as graph edges, and the transaction frequency data as the weight attributes of the graph edges, a local transaction graph is constructed. Extract the topological structure information of the local transaction graph to generate the transaction flow graph data.

4. The smart contract-based anti-counterfeiting information verification system for commodity coupons according to claim 3, characterized in that, The method for analyzing the transaction flow graph data using a graph neural network model to generate risk exposure assessment results includes: Input the transaction flow graph data into the graph neural network model; The graph convolutional layer in the graph neural network model is used to perform node feature aggregation and graph pooling operations on the transaction flow graph data to extract local anomaly features and global topological features of the transaction flow graph data. The local anomaly features and the global topological features are concatenated to generate a comprehensive feature vector; The comprehensive feature vector is input into the preset fully connected classification layer in the graph neural network model to calculate the fraud probability value, and the fraud probability value is mapped to the risk exposure assessment result.

5. A smart contract-based anti-counterfeiting information verification system for commodity vouchers according to claim 2, characterized in that, The methods for triggering the dynamic routing strategy of the smart contract and outputting the target verification level include: Obtain the preset danger threshold and preset time threshold; The consensus friction entropy index is compared with the danger threshold, and the safe time window is compared with the preset time threshold. If the consensus friction entropy index is less than the danger threshold and the safety time window is greater than or equal to the preset time threshold, then a regular verification level is assigned according to the risk exposure assessment result, and the regular verification level is used as the target write-off verification level. If the consensus friction entropy index is greater than or equal to the danger threshold, or the security time window is less than the preset time threshold, then a degradation routing rule is triggered, and the target verification level is output based on the degradation routing rule.

6. A smart contract-based anti-counterfeiting information verification system for commodity coupons according to claim 5, characterized in that, The method for outputting the target verification level based on the downgrade routing rules includes: Obtain a preset controllable risk threshold; Compare the risk exposure assessment results with the controllable risk threshold; If the risk exposure assessment result is less than the controllable risk threshold, a lightweight verification level is generated, and the lightweight verification level is used as the target write-off verification level. If the risk exposure assessment result is greater than or equal to the controllable risk threshold, a delay processing level is generated, and the delay processing level is used as the target write-off verification level.

7. A smart contract-based anti-counterfeiting information verification system for commodity coupons according to claim 6, characterized in that, The method for executing the corresponding reconciliation verification logic based on the target reconciliation verification level includes: If the target verification level is the lightweight verification level, the basic signature verification logic is triggered, the full-link tracing operation is stopped, and an execution result indicating that the verification has passed is generated. If the target verification level is the delayed processing level, the coupon verification request is suspended and moved to a preset delayed queue, the current verification operation is stopped, and an execution result indicating that the request is suspended is generated. If the target verification level is the regular verification level, the full-link traceability verification logic is triggered to perform in-depth verification of the issuance and circulation nodes of the voucher and generate the corresponding execution result.

8. A smart contract-based anti-counterfeiting information verification system for commodity coupons according to claim 1, characterized in that, The method for adjusting the resource consumption parameters of the smart contract based on the write-off status to provide feedback control over the consensus friction entropy index of the blockchain network includes: Obtain the basic execution computing power overhead parameters of the smart contract; The risk exposure assessment result and the consensus friction entropy index are linearly normalized and then multiplied to obtain the comprehensive risk value. The comprehensive risk value is mapped to a dynamic penalty coefficient based on a preset positive correlation mapping function; Multiply the basic execution computing power overhead parameter by the dynamic penalty coefficient to obtain the target execution computing power overhead parameter, which serves as the resource consumption parameter; The resource consumption parameters are broadcast to the blockchain network to reduce the number of concurrent requests by increasing the computing power overhead of abnormal transactions, thereby providing feedback control over the consensus friction entropy index.

Images