An asynchronous consensus method for blockchain based on GRA-HB-BFT

Abstract

This invention provides a blockchain asynchronous consensus method based on GRA-HB-BFT, belonging to the field of blockchain asynchronous consensus technology. This invention employs grey relational analysis combined with entropy method to construct a dynamic node reputation evaluation model, objectively evaluating consensus node performance through multi-dimensional indicators and outputting a comprehensive reputation value. Secondly, the traditional asynchronous common subset protocol is decoupled into a reliable broadcast protocol area and an asynchronous Byzantine protocol area, enabling parallel execution of the two areas. Consensus nodes perform threshold encryption on transaction proposals and broadcast them. Then, based on the reputation evaluation results, broadcast instances of high-reputation consensus nodes are selected as input for Byzantine consensus, effectively reducing communication overhead. Finally, after parallel processing, the encrypted proposals are threshold decrypted, verified, and sorted before being submitted to the ledger. Simultaneously, the comprehensive reputation value is dynamically adjusted based on the consensus node performance, forming a continuously optimized closed-loop control mechanism.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of blockchain asynchronous consensus, in particular to a blockchain asynchronous consensus method based on GRA-HB-BFT. BACKGROUND

[0002] In the field of blockchain technology, asynchronous consensus algorithm has important value in distributed systems due to its strong robustness of not relying on network timing assumptions. HB-BFT (HoneyBadgerBFT) as a typical asynchronous consensus protocol integrates reliable broadcast and asynchronous binary Byzantine protocol through asynchronous public subset protocol to achieve secure consensus in the presence of malicious consensus nodes. However, with the expansion of consensus node size and the complexity of network environment, the existing HB-BFT protocol faces severe challenges in practical application: the fixed consensus node participation mechanism and static instance processing flow of the existing HB-BFT protocol lead to a significant decrease in throughput in high concurrency scenarios, and lack of adaptive ability when facing malicious consensus nodes or performance unstable consensus nodes, which seriously affects the consensus efficiency and system scalability.

[0003] In the prior art, improvements to the HB-BFT protocol are mostly focused on optimizing network transmission or parallel processing structure, for example, by logically separating the reliable broadcast protocol in the asynchronous public subset protocol from the asynchronous binary Byzantine protocol stage to achieve partial parallel execution, or introducing a simple reputation mechanism based on historical behavior to filter consensus nodes participating in consensus. These methods improve consensus efficiency to some extent, but their consensus node selection mechanism relies on fixed thresholds or static evaluation, which cannot reflect the dynamic performance changes of consensus nodes in real time. At the same time, there is a lack of multi-dimensional comprehensive evaluation and continuous tracking mechanism for the behavior of consensus nodes, which leads to insufficient recognition accuracy when facing intentionally inefficient consensus nodes or intermittent malicious behavior, affecting the overall consensus stability.

[0004] Although the existing improvement scheme attempts to improve performance through parallelization and basic screening, its core deficiency is the inability to establish a dynamic, accurate and self-adaptable consensus node reputation evaluation and selection system. Specifically, the existing technology cannot effectively solve the following problems: First, the selection of consensus nodes lacks objective quantification and fusion evaluation of multi-dimensional performance indicators such as response timeliness, transaction processing stability and historical participation, resulting in limited optimization accuracy. Second, there is a lack of real-time reward and punishment feedback mechanism in the consensus process, which cannot form continuous inhibition of malicious consensus nodes and cannot timely encourage high-performance consensus nodes. Finally, the stage coupling and static instance processing mode of the traditional asynchronous public subset protocol lead to insufficient resource utilization, making it difficult to achieve continuous and efficient consensus throughput in complex network environments.

[0005] The above information disclosed in the background section is only for the purpose of enhancing the understanding of the background of the present disclosure, and therefore it can include information that does not constitute the prior art known to those of ordinary skill in the art. SUMMARY

[0006] The purpose of the present application is to provide a blockchain asynchronous consensus method based on GRA-HB-BFT to solve the problems raised in the background.

[0007] To achieve the above purpose, the present application provides the following technical solutions:

[0008] A blockchain asynchronous consensus method based on GRA-HB-BFT, the specific steps include:

[0009] Step 1: Adopting grey correlation analysis combined with entropy method, a consensus node reputation evaluation model is constructed, which uses entropy method to calculate the information entropy of each evaluation index to determine its objective weight, and through grey correlation degree calculation, the correlation of the evaluation index sequence of each consensus node with the ideal reference sequence is evaluated, and finally the comprehensive reputation value of each consensus node is output;

[0010] Step 2: Decouple the asynchronous public subset protocol into two parallel execution processing areas, area one is a reliable broadcast protocol area, and area two is an asynchronous Byzantine protocol area, in area one, after each consensus node receives the transaction proposal submitted by the client, it performs threshold encryption processing on it, and generates the encrypted proposal as a reliable broadcast protocol instance for broadcast;

[0011] Step 3: In area two, based on the pre-constructed consensus node reputation evaluation model, select the instance initiated by the high-reputation consensus node from the multiple reliable broadcast protocol instances broadcast from area one as input, which is used for subsequent consensus decision of asynchronous Byzantine protocol;

[0012] Step 4: After parallel processing of the reliable broadcast protocol and the asynchronous Byzantine protocol, the encrypted proposal reached consensus is threshold decrypted, correctness verified and timestamp sorted, and submitted to the distributed ledger, and at the same time, after each round of consensus is completed, according to the behavior performance of the consensus node in this round of consensus, its comprehensive reputation value is adjusted through a dynamic reward and punishment mechanism.

[0013] Further, the construction of the node reputation evaluation model specifically includes:

[0014] The response time of the consensus node, the transaction processing success rate, and the historical consensus participation degree are selected as the evaluation indexes, wherein the response time refers to the time used from receiving a transaction request by the consensus node to completing the processing and broadcasting the response to the network; the transaction processing success rate refers to the ratio of the number of transactions successfully processed and submitted to the ledger by the consensus node in a unit of time to the total number of received transactions; the historical consensus participation degree refers to the proportion of the consensus node actively participating in and completing the consensus process in the past N rounds of consensus, and N is a preset historical round threshold;

[0015] Behavior data of each consensus node in the historical rounds is collected to form an evaluation data set, and the evaluation data set is standardized to eliminate the dimensional differences between different evaluation indexes;

[0016] Based on the standardized data, the information entropy of each evaluation index is calculated by using the entropy method, and the objective weight of each evaluation index is determined by using the entropy method according to the information entropy; the optimal value of each evaluation index in all consensus nodes is selected to form an ideal reference sequence; the optimal value is determined according to the evaluation index, and the maximum value is taken for a positive evaluation index, and the minimum value is taken for a negative evaluation index;

[0017] The grey correlation coefficient between the evaluation index value of each consensus node and the corresponding index of the ideal reference sequence is calculated, and the comprehensive reputation value of each consensus node is calculated by comprehensively considering the grey correlation coefficient and the objective weight of each evaluation index.

[0018] Further, the data standardization processing adopts the range standardization method, specifically:

[0019] For each evaluation index, the maximum value and the minimum value of the evaluation index in the evaluation data set of all consensus nodes are identified, and the evaluation data of each consensus node in the evaluation index is scaled to the interval [0, 1] according to the maximum value and the minimum value;

[0020] The information entropy of each evaluation index is calculated by using the entropy method, and the process is as follows: based on the standard value after the standardization processing, the proportion of the standard value of each consensus node in each evaluation index to the sum of all standard values of the evaluation index is calculated, and the information entropy of each evaluation index is calculated according to the proportion distribution by using the information entropy theory;

[0021] The objective weight of each evaluation index is determined, and the process is as follows: according to the information entropy of each evaluation index, the information utility value is calculated, the information utility value of each evaluation index is compared with the sum of the information utility values of all evaluation indexes, and the objective weight of each evaluation index is determined.

[0022] Further, the optimal value of each evaluation index is selected to form an ideal reference sequence , wherein,​ For the ideal reference sequence in the 1st The optimal value on each evaluation metric; for the i-th consensus node, its standardized data sequence is: ,in Let j be the standard value of the j-th evaluation metric in the i-th consensus node, where j is the index of the evaluation metric;

[0023] Calculate the relationship between the i-th consensus node and the ideal reference sequence at the i-th... The formula for the grey relational coefficient on each evaluation indicator is:

[0024] ;

[0025] in, Let $\frac{j}{i}$ be the absolute deviation between the standard value and the ideal value of the $j$-th evaluation metric in the $i$-th consensus node. and These represent the minimum and maximum absolute deviations of the standard and ideal values ​​for all consensus nodes across all evaluation metrics, respectively. The resolution coefficient is used to adjust the resolution of the gray relational coefficient. For the first The consensus node is at the _th ... Grey relational coefficients on each evaluation indicator;

[0026] Calculate the first The overall reputation value of each consensus node Use the following formula:

[0027] ;

[0028] in, For the first The objective weights of each evaluation indicator, This is used for subsequent sorting and selection of consensus nodes.

[0029] Furthermore, instances initiated by high-reputation consensus nodes are preferred as input, employing a two-phase strategy:

[0030] Before running the system The consensus round is defined as the initialization phase, in which a random strategy is used to select the consensus round. Several reliable broadcast protocol instances participate in the asynchronous Byzantine protocol consensus, among which The preset round threshold, This represents the maximum number of fault-tolerant consensus nodes in the system.

[0031] After completion After the consensus round, the system enters a stable operation phase. During this phase, the nodes are ranked according to their latest comprehensive reputation scores, and the nodes with the highest comprehensive reputation scores are selected. reliable broadcast protocol instance published by the consensus node participates in the asynchronous Byzantine protocol consensus;

[0032] The node reputation evaluation model performs a comprehensive recalculation every round of consensus, in which is the preset update period; between two comprehensive updates, the comprehensive reputation value of the consensus node is adjusted in a small range through a dynamic reward and punishment mechanism.

[0033] Further, when processing the transaction proposal in the reliable broadcast protocol area, the consensus node performs the following operations:

[0034] Randomly extract a preset number of transaction proposals from the locally maintained transaction pool, and use a threshold encryption algorithm to encrypt the batch of transaction proposals; the ciphertext formed after encryption is used as the input content of the reliable broadcast protocol instance of the consensus node, and is broadcast to all other consensus nodes in the network;

[0035] The specific process of threshold decryption of the encrypted proposal is:

[0036] When the system reaches consensus on a batch of encrypted proposals through the asynchronous Byzantine protocol, each consensus node begins to collect decryption fragments from other consensus nodes; when any consensus node collects not less than valid decryption fragments, the plaintext transaction is reconstructed using these fragments; the decrypted transactions are subjected to redundancy elimination and logical correctness verification; all transactions that pass the verification are globally sorted according to their timestamps and packaged into official blocks, which are finally submitted to the distributed ledger for persistent storage.

[0037] Further, the dynamic adjustment of the comprehensive reputation value is achieved through a dynamic reward and punishment mechanism, specifically:

[0038] Immediately after each round of consensus, the behavior of the consensus node is verified; if the consensus node provides an effective reliable broadcast protocol instance in this round of consensus, a preset reward value is added to its original comprehensive reputation value; if the consensus node provides an invalid reliable broadcast protocol instance, a preset penalty value is subtracted from its original comprehensive reputation value;

[0039] If the comprehensive reputation value of the consensus node fails to enter the top in the updated ranking, the reliable broadcast protocol instance initiated by the consensus node will no longer be selected in subsequent consensus;

[0040] The parallel execution mechanism of the region one and the region two is realized by time overlap: when the part reliable broadcast protocol instance in the region one completes the broadcast, the region two immediately starts to process the consensus of these instances, without waiting for all instances to complete the broadcast, while the region one continues to process the broadcast of the remaining instances.

[0041] Compared with the prior art, the present application has the beneficial effects that:

[0042] The present application effectively solves the problems of static selection mechanism and single evaluation dimension of the consensus node in the prior art by introducing a dynamic node reputation evaluation model based on grey correlation analysis and entropy method, the multi-dimensional evaluation index adopted, including response time, transaction processing success rate and historical participation degree, and the objective weight determined by the entropy method, can accurately quantify the real-time performance and reliability of the consensus node; and the grey correlation degree calculation realizes dynamic evaluation of the comprehensive reputation of the node by comparing the similarity of the consensus node sequence and the ideal reference sequence, which enables the system to accurately identify and preferentially select high-performance consensus nodes to participate in the key consensus phase, thereby improving the quality of the consensus group from the source and significantly enhancing the identification ability and system robustness of malicious consensus nodes and inefficient consensus nodes.

[0043] The present application overcomes the resource idling and efficiency bottleneck caused by the phase coupling of the original protocol by decoupling the traditional asynchronous common subset protocol into a reliable broadcast protocol region and an asynchronous Byzantine protocol region and realizing parallel execution, optimizes the broadcast instance of the consensus node with high reputation based on the dynamic reputation evaluation result to participate in the Byzantine consensus, greatly reduces the number of consensus instances that need to be processed, thereby significantly reducing the network communication overhead, and the combination mechanism of protocol decoupling and instance optimization enables the broadcast process and the consensus process to fully overlap in the time dimension, the system resources are efficiently utilized, and finally the system can still maintain high throughput and low delay in a complex network environment.

[0044] The dynamic reward and punishment mechanism and the update of the comprehensive reputation value after each round of consensus established by the present application form a continuous optimization closed loop, which can adjust the comprehensive reputation value of the consensus node in real time according to its latest behavior: the consensus node providing effective instances is rewarded, and the consensus node with abnormal behavior is punished and its participation priority is timely reduced, this feature enables the system to have adaptive ability, can continuously optimize the node selection strategy without restarting or manual intervention, suppress the influence of malicious behavior, thereby maintaining stable high performance and strong robustness in the long-term operation, and effectively solves the performance degradation problem caused by the lack of continuous optimization mechanism in the prior art. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 It is a schematic diagram of the overall method flow of the present application;

[0046] Figure 2This is a scatter plot of the grey relational coefficient and comprehensive reputation value of this invention.

[0047] Figure 3 This is a bar chart showing the grey relational coefficient, objective weight, and comprehensive reputation value of this invention.

[0048] Figure 4 This is a fitted curve of the grey relational coefficient and the comprehensive reputation value of this invention;

[0049] Figure 5 This is a curve showing the fitting of the objective weight-comprehensive reputation value of this invention. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0051] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0052] Example:

[0053] Please see Figures 1-5 The present invention provides a technical solution:

[0054] An asynchronous consensus method for blockchain based on GRA-HB-BFT, comprising the following steps:

[0055] Step 1: A consensus node reputation evaluation model is constructed by combining grey relational analysis with entropy method. The entropy method is used to calculate the information entropy of each evaluation indicator to determine its objective weight. The grey relational degree is calculated to evaluate the correlation between the evaluation indicator sequence of each consensus node and the ideal reference sequence. Finally, the comprehensive reputation value of each consensus node is output.

[0056] The construction of a node reputation assessment model specifically includes:

[0057] The response time of consensus nodes, transaction processing success rate, and historical consensus participation are selected as evaluation indicators. Response time refers to the time taken from when a consensus node receives a transaction request to when it completes processing and broadcasts the response to the network. Transaction processing success rate refers to the ratio of the number of transactions successfully processed and committed to the ledger by a consensus node per unit time to the total number of transactions received. Historical consensus participation refers to the proportion of consensus nodes that actively participate in and complete the consensus process in the past N rounds of consensus, where N is a preset threshold for the number of historical rounds.

[0058] In specific implementations, considering the diversity of consensus node performance and behavioral characteristics in the asynchronous consensus environment of blockchain, different consensus nodes have significant differences in response speed, processing capacity and participation enthusiasm. By establishing a multi-dimensional node reputation evaluation system, the comprehensive performance of each consensus node can be quantified in a comprehensive and objective manner, providing a reliable basis for the selection of consensus nodes in the subsequent consensus process, thereby improving the overall consensus efficiency and security of the system.

[0059] Response time is a precise measurement of the time interval from when a consensus node receives a complete transaction data packet from its network interface to when the consensus node completes transaction verification, packaging, and successfully broadcasts the transaction response data packet to its peer consensus nodes. Transaction processing success rate is the proportion of transactions that a consensus node correctly processes and that are ultimately included in the blockchain within a statistical period (e.g., the most recent 1000 block generation cycles). Historical consensus participation is the percentage of consensus nodes that actually complete the entire consensus process in the most recent N rounds (e.g., N is 500). Rounds in which a node is absent or withdraws midway are not counted as valid participation.

[0060] The reason for choosing consensus node response time, transaction processing success rate, and historical consensus participation as core evaluation indicators is that these three evaluation indicators can accurately reflect the operating status and reliability of consensus nodes from different dimensions. Response time directly reflects the processing efficiency of consensus nodes, transaction processing success rate reflects the work quality of consensus nodes, and historical consensus participation characterizes the stability and willingness to cooperate of consensus nodes. The combined evaluation of these three evaluation indicators can effectively identify high-performance consensus nodes while filtering out malicious or low-performance consensus nodes.

[0061] A distributed data collection mechanism is established, where each consensus node records the basic data of the above evaluation indicators locally. After each round of consensus, the data is aggregated to the evaluation system through a secure channel. The original data to be evaluated is verified for integrity and then constructed into a two-dimensional evaluation matrix, with rows corresponding to consensus nodes and columns corresponding to evaluation indicators.

[0062] Collect behavioral data of each consensus node in historical rounds to form an evaluation dataset, and standardize the evaluation dataset to eliminate the differences in the units of measurement between different evaluation indicators.

[0063] Based on the standardized data, the information entropy value of each evaluation indicator is calculated using the entropy method. According to the information entropy value, the objective weight of each evaluation indicator is determined using the entropy method. The optimal value of each evaluation indicator in all consensus nodes is selected to form an ideal reference sequence. The optimal value is determined according to the evaluation indicator, with the maximum value taken for positive evaluation indicators and the minimum value taken for negative evaluation indicators.

[0064] Calculate the grey relational coefficient between each evaluation index value of each consensus node and the corresponding index of the ideal reference sequence. Combine the grey relational coefficient with the objective weight of each evaluation index to calculate the comprehensive reputation value of each consensus node.

[0065] Data standardization is performed using the range standardization method, specifically:

[0066] For each evaluation metric, identify the maximum and minimum values ​​of that metric across all consensus nodes' datasets to be evaluated. Based on these maximum and minimum values, scale the data to be evaluated for each consensus node according to that metric. Within the range;

[0067] The information entropy value of each evaluation indicator is calculated using the entropy method. The process is as follows: Based on the standardized values, the proportion of the standard value of each consensus node on each evaluation indicator to the sum of all standard values ​​of that evaluation indicator is calculated. Based on the distribution of this proportion, the information entropy value of each evaluation indicator is calculated using the information entropy theory.

[0068] It should be noted that the data standardization process employs range standardization, the core purpose of which is to eliminate differences in the dimensions and numerical ranges of different evaluation indicators, making the data of each evaluation indicator comparable. In the specific implementation process, the maximum and minimum values ​​of all consensus node data are calculated for each evaluation indicator, and then the original data of each consensus node is linearly transformed proportionally to the range standardization method. Within the interval, this process ensures that subsequent weight calculations and correlation analyses are based on a unified numerical foundation;

[0069] The objective weight of each evaluation indicator is determined as follows: based on the information entropy value of each evaluation indicator, its information utility value is calculated, and the information utility value of each evaluation indicator is compared with the sum of the information utility values ​​of all evaluation indicators to determine the objective weight of each evaluation indicator.

[0070] In a specific implementation, the calculation of information utility value adopts a deterministic mathematical definition, namely, information utility value equals 1 minus information entropy value. The practical significance of this relationship is that information entropy value directly reflects the degree of disorder or unreliability of the corresponding evaluation index data. When the information entropy value of a certain evaluation index is larger, it indicates that the credibility of the evaluation index in distinguishing the performance of consensus nodes is lower, and its ability to provide effective discriminative information is also weaker. Objective weight is proportional to information utility value, so the larger the information entropy value, the smaller the information utility value and the smaller the objective weight.

[0071] Regarding weight determination, the entropy method is used for objective weighting. This is because, compared to subjective weighting methods, the entropy method can automatically determine the importance of each evaluation indicator based on the dispersion of the data itself. In practice, the weight distribution of each consensus node value under each evaluation indicator is first calculated. Then, the information entropy value of each evaluation indicator is calculated based on the information entropy theory. The smaller the information entropy, the greater the role of the evaluation indicator in distinguishing the performance of consensus nodes, and the higher the weight is assigned accordingly. This method effectively avoids the bias caused by subjective judgment and ensures the objectivity and accuracy of the evaluation results.

[0072] Furthermore, the optimal values ​​of each evaluation indicator are selected to form an ideal reference sequence. ,in, For the ideal reference sequence in the 1st The optimal value on each evaluation metric; for the i-th consensus node, its standardized data sequence is: ,in Let j be the standard value of the j-th evaluation metric in the i-th consensus node, where j is the index of the evaluation metric;

[0073] when At that time, it represents the response time of the consensus node, which is an evaluation metric that quantifies the processing efficiency of the consensus node, and its optimal value is... The minimum value among all consensus nodes is standardized, because response time is a negative indicator that is expected to be as short as possible;

[0074] when At that time, it represents the transaction processing success rate, an evaluation metric that measures the working quality and reliability of consensus nodes, with its optimal value being... The maximum value in the data is standardized across all consensus nodes, because the success rate is a positive metric that is expected to be as high as possible;

[0075] when At that time, it represents the evaluation index of historical consensus participation, which reflects the continued participation willingness and stability of consensus nodes, and its optimal value is... The maximum value in the standardized data of all consensus nodes is used because participation is also a positive indicator that the higher the expectation, the better;

[0076] In a specific implementation, the node reputation evaluation model is constructed using the analysis method of grey system theory. It quantifies the degree of closeness of each consensus node to the ideal state by establishing an ideal reference sequence and calculating grey relational coefficients.

[0077] When constructing the ideal reference sequence, the optimal values ​​for the three evaluation indicators are selected to form the reference benchmark. Among them, response time is used as a negative indicator, and the minimum standardized value among all consensus nodes is selected; transaction success rate and historical consensus participation are used as positive indicators, and the maximum standardized value among all consensus nodes is selected. The ideal reference sequence thus constitutes the theoretically optimal performance state of consensus nodes, providing a scientific reference standard for subsequent correlation comparison.

[0078] Calculate the relationship between the i-th consensus node and the ideal reference sequence at the i-th... The formula for the grey relational coefficient on each evaluation indicator is:

[0079] ;

[0080] in, Let $\frac{j}{i}$ be the absolute deviation between the standard value and the ideal value of the $j$-th evaluation metric in the $i$-th consensus node. and These represent the minimum and maximum absolute deviations of the standard and ideal values ​​for all consensus nodes across all evaluation metrics, respectively. The resolution coefficient is used to adjust the resolution of the gray relational coefficient. For the first The consensus node is at the _th ... Grey relational coefficients on each evaluation indicator;

[0081] The calculation of the grey relational coefficient is a key step in the evaluation process. This coefficient reflects how close a single consensus node's performance on a specific evaluation metric is to the ideal state. The absolute deviation term in the calculation formula directly reflects the gap between the current metric value and the ideal value of the consensus node. The closer the performance of the consensus node is to the ideal state, the smaller the deviation value. The global minimum deviation and global maximum deviation in the formula provide a reference for the performance fluctuation range of the entire system, ensuring that the calculation results have system-level comparability.

[0082] It should be noted that the resolution coefficient The setting of this parameter has a significant impact on the sensitivity of the correlation coefficient. In practical applications, extensive experimental verification has shown that... Setting the value to 0.5 can maintain sufficient distinguishability while avoiding oversensitivity. This value is based on in-depth research on the performance fluctuation characteristics of consensus nodes in actual blockchain network environments, and can effectively balance the system's response speed and stability requirements to changes in consensus node performance.

[0083] Calculate the first The overall reputation value of each consensus node Use the following formula:

[0084] ;

[0085] in, For the first The objective weights of each evaluation indicator, Used for subsequent sorting and selection of consensus nodes;

[0086] After calculating the grey relational coefficients of each indicator, the weights of the indicators determined by the entropy method in the early stage are combined and weighted to obtain the comprehensive reputation value of each consensus node. This calculation process fully considers the differences in the importance of different evaluation indicators in the performance evaluation of consensus nodes, ensuring the scientificity and comprehensiveness of the evaluation results. The level of the comprehensive reputation value directly reflects the overall performance of the consensus node. The higher the value, the closer the comprehensive performance of the consensus node is to the ideal state.

[0087] Specifically, the grey relational coefficient The range of values ​​is Its numerical value directly reflects how close the performance of the i-th consensus node to the ideal state on the j-th evaluation metric. A larger value indicates that the consensus node performs better on this evaluation metric and the smaller the gap from the ideal value. This grey relational coefficient is related to the absolute deviation. There is a negative correlation, meaning that the closer the actual performance of the consensus node is to the ideal value and the smaller the absolute deviation, the larger the corresponding grey correlation coefficient.

[0088] Overall credit score As the final evaluation result, its value directly determines the priority of consensus nodes in the subsequent instance selection process. The larger the value, the better the overall performance of the consensus node, and the higher the probability of it being selected in the consensus process. This mechanism ensures that high-performance consensus nodes can get more opportunities to participate, thereby improving the operating efficiency and reliability of the entire system.

[0089] The specific data of some node numbers and comprehensive reputation values ​​are shown in Table 1.

[0090] Table 1. Comprehensive Credit Score Statistics

[0091]

[0092] Analysis of the data revealed a clear correlation between the node evaluation parameters. The data showed a stable negative correlation between absolute deviation and the grey correlation coefficient. As the node number increased, the absolute deviation fluctuated, while the grey correlation coefficient exhibited the opposite trend. When the absolute deviation was small, such as when the absolute deviation of consensus node 3 was 0.08, the corresponding grey correlation coefficient reached 0.87. However, when the absolute deviation increased, such as when the absolute deviation of consensus node 7 was 0.52, the grey correlation coefficient dropped to 0.49. This indicates that the closer the actual performance of a consensus node is to the ideal reference sequence, the higher its grey correlation coefficient.

[0093] When analyzing the relationship between the overall reputation score and various parameters, the grey relational coefficient has a significant impact on the overall reputation score. When the grey relational coefficient is higher, the overall reputation score of the consensus node is also higher. For example, the grey relational coefficient of consensus node 14 is 0.90, and its overall reputation score reaches 0.98; while the grey relational coefficient of consensus node 19 is 0.47, and its overall reputation score is only 0.48. This relationship shows that the grey relational coefficient is a key factor in determining the reputation rating of consensus nodes.

[0094] Meanwhile, the distribution of objective weights also has a significant impact on the final comprehensive reputation value. There are certain differences in objective weights among different consensus nodes. These differences reflect the importance of each evaluation indicator in the reputation evaluation system. Indicators with higher objective weights contribute more to the final reputation value. However, the fundamental role of the gray relation coefficient is always decisive. For example, consensus node 3 and consensus node 8 both have high gray relation coefficients. Although their objective weights are different, they both ultimately obtained high comprehensive reputation values.

[0095] From the overall data trend, the comprehensive reputation value of consensus nodes is mainly dominated by the gray correlation coefficient, while the objective weight plays an appropriate adjustment role. This evaluation mechanism can objectively reflect the actual performance of consensus nodes and provide a reliable basis for subsequent consensus node selection. In practical applications, the combined effects of absolute deviation, gray correlation coefficient and objective weight should be considered to optimize the consensus node selection strategy and improve the overall performance of the blockchain system.

[0096] Step 2: Decouple the asynchronous common subset protocol into two parallel processing areas. Area 1 is the reliable broadcast protocol area, and Area 2 is the asynchronous Byzantine protocol area. In Area 1, after each consensus node receives the transaction proposal submitted by the client, it performs threshold encryption processing on it and generates an encrypted proposal as a reliable broadcast protocol instance for broadcasting.

[0097] When processing transaction proposals in the reliable broadcast protocol area, consensus nodes perform the following operations:

[0098] A predetermined number of transaction proposals are randomly selected from the locally maintained transaction pool, and these proposals are encrypted using a threshold encryption algorithm. The encrypted ciphertext proposals are then used as input to the reliable broadcast protocol instance of this consensus node and broadcast to all other consensus nodes in the network.

[0099] In a specific implementation, significant architectural optimizations were made to the traditional asynchronous common subset protocol, decoupling it into two parallel processing areas. This design is based on our in-depth analysis of the blockchain consensus process, which revealed that in the traditional serial execution mode, consensus nodes need to wait for all instances to be ready before they can begin consensus after completing the broadcast, resulting in a large amount of idle computing resources and wasted time.

[0100] During the execution of the reliable broadcast protocol in Region 1, each consensus node first randomly selects a specific number of transaction proposals from its locally maintained transaction pool. This random selection mechanism ensures the fairness of transaction processing and prevents attacks launched by malicious consensus nodes through specific transaction ordering. In practice, consensus nodes use... Threshold encryption algorithms encrypt selected batches of transactions, where... This represents the minimum number of consensus nodes required for decryption. This represents the total number of consensus nodes in the system. This encryption method ensures that even if some consensus nodes are maliciously controlled, they cannot decrypt the transaction content individually, thereby maintaining the security of the system.

[0101] After encryption, the consensus node encapsulates the generated ciphertext proposal into a standard format of a reliable broadcast protocol instance. This format contains necessary metadata information, such as transaction hash, timestamp, and digital signature. Subsequently, the consensus node broadcasts its instance to all other consensus nodes through a peer-to-peer network. During the broadcast, a Merkle tree structure is used to verify the integrity of the data. Each intermediate consensus node verifies the validity of the message before forwarding it, ensuring that malicious consensus nodes cannot tamper with the broadcast content.

[0102] Meanwhile, the asynchronous Byzantine protocol region of Region 2 does not need to wait for all broadcast instances to complete. Instead, it adopts a smart triggering mechanism. When the system detects that a sufficient number of high-quality broadcast instances are ready, it will immediately start the consensus process. This design enables the two regions to achieve true parallel execution, significantly improving the overall throughput of the system.

[0103] In actual deployment, a dynamic load balancing mechanism is also introduced. This system can monitor the execution status of the two regions in real time and dynamically adjust resource allocation according to network conditions and consensus node performance. For example, when network bandwidth is sufficient, the system will appropriately increase the concurrency of the broadcast region; while when the network is congested, the resource needs of the consensus region will be prioritized. This fine-grained resource management ensures that the system can maintain optimal performance under various working conditions.

[0104] In a specific implementation, the protocol decoupling is not a simple logical division, but rather the construction of two independent processing engines at the system architecture level, which are managed by a unified consensus state coordinator.

[0105] It is worth noting that Region 1 is a reliable broadcast protocol processing engine. This engine includes a transaction collection module, an encryption module, a broadcast control module, and an instance cache library. Its workflow is as follows: the transaction collection module obtains transactions from the local transaction pool according to a preset strategy (such as random selection); the encryption module uses... The threshold encryption algorithm is used for processing; the broadcast control module is responsible for encapsulating the ciphertext into a standard reliable broadcast protocol instance format and managing its propagation process throughout the network to ensure message reliability and consistency; all broadcast instance copies are stored in the instance cache library for other components to query.

[0106] Region 2 is the asynchronous Byzantine protocol processing engine. This engine includes an instance selection module, a voting consensus module, and a consensus result storage. Its workflow is as follows: The instance selection module continuously monitors the instance cache based on the consensus node reputation evaluation model and selects asynchronous Byzantine protocol instances from high-reputation consensus nodes; The voting consensus module organizes consensus nodes to conduct multiple rounds of voting on these selected instance inputs and executes the binary Byzantine consensus algorithm; The consensus result reached, i.e., which instances are finally confirmed and stored in the consensus result storage;

[0107] The parallel execution of the two engines is coordinated by a consensus process scheduler, which maintains a global consensus round state. Once the reliable broadcast protocol processing engine stores the first batch of eligible instances in the cache, the consensus process scheduler will immediately trigger the asynchronous Byzantine protocol processing engine to start working, without waiting for all consensus nodes' reliable broadcast protocol instances to complete broadcasting.

[0108] While the asynchronous Byzantine protocol engine reaches consensus on the first batch of instances, the reliable broadcast protocol engine continues to process the broadcast tasks of the remaining consensus nodes in parallel. New instances generated subsequently can be incrementally incorporated into the consensus process by the asynchronous Byzantine protocol engine. The two engines communicate and exchange data through a shared instance cache library and consensus state coordinator to ensure that the processing progress is both independent and coordinated.

[0109] Step 3: In Region 2, based on the pre-built consensus node reputation evaluation model, select instances initiated by high-reputation consensus nodes from multiple reliable broadcast protocol instances broadcast in Region 1 as input for subsequent consensus decisions in the asynchronous Byzantine protocol.

[0110] Instances initiated by high-reputation consensus nodes are preferred as input, employing a two-phase strategy:

[0111] Before running the system The consensus round is defined as the initialization phase, in which a random strategy is used to select the consensus round. Several reliable broadcast protocol instances participate in the asynchronous Byzantine protocol consensus, among which The preset round threshold, This represents the maximum number of fault-tolerant consensus nodes in the system.

[0112] After completion After the consensus round, the system enters a stable operation phase. During this phase, the nodes are ranked according to their latest comprehensive reputation scores, and the nodes with the highest comprehensive reputation scores are selected. The reliable broadcast protocol instances published by the consensus nodes participate in the asynchronous Byzantine protocol consensus;

[0113] Node reputation assessment model every A consensus round will be used to conduct a comprehensive recalculation, in which... The update cycle is preset; between two full updates, the overall reputation value of the consensus nodes is adjusted within a small range through a dynamic reward and punishment mechanism.

[0114] In a specific implementation, the instance selection mechanism in Region 2 is a dynamic and adaptive process. Its core lies in balancing the exploration needs at the initial stage of system startup with the performance optimization during the stable operation period through a two-stage strategy.

[0115] The system will be forward The consensus round is set as the initialization phase. This is based on a deep understanding of the dynamic characteristics of network consensus. In this phase, since sufficient consensus node behavior data has not yet been accumulated to establish a reliable reputation assessment, a random strategy is used to select from all broadcast instances. Each instance participates in the asynchronous Byzantine Fault Tolerance (BFT) consensus mechanism. This design ensures that newly joined and unknown consensus nodes have an equal opportunity to participate in the consensus process, avoiding the problem of high-quality consensus nodes being overlooked due to insufficient initial reputation data. (Parameters) The setting needs to take into account the network size, consensus node stability and system convergence speed, and the optimal value is usually determined through experimental analysis.

[0116] After the initialization phase, the system enters the stable operation phase. At this point, the node reputation evaluation model has accumulated sufficient historical data to accurately reflect the true performance level of each consensus node. During this phase, the system strictly sorts nodes according to their latest comprehensive reputation values ​​and selects the top-ranked nodes. Consensus node instances participate in consensus, and this mechanism ensures that consensus resources are always tilted towards high-performance and high-reliability consensus nodes, thereby significantly improving the overall efficiency and security of the system.

[0117] It is important to note that the instance selection process and the updating and maintenance of the credit assessment model form a tightly integrated closed-loop system. The credit assessment model is updated and maintained every [period]. Each consensus round involves a comprehensive recalculation. This update cycle balances the trade-off between model accuracy and computational overhead. Shorter update cycles reflect changes in consensus node performance more quickly but increase the system's computational burden; longer update cycles, on the other hand, are dynamically adjusted by monitoring system load and network conditions in actual deployments. Values ​​are set to achieve optimal system performance.

[0118] Between the two comprehensive updates, the system uses a dynamic reward and punishment mechanism to fine-tune the overall reputation value of consensus nodes. This mechanism can respond to changes in the behavior of consensus nodes in real time: positive incentives are given to consensus nodes that continuously provide high-quality services, so that their overall reputation value can be steadily improved; and corresponding penalties are implemented for consensus nodes that exhibit abnormal behavior, so that their reputation rating can be reduced in a timely manner. This instant feedback mechanism ensures the timeliness and accuracy of reputation assessment, enabling the system to quickly adapt to changes in the network environment.

[0119] At the technical implementation level, the instance selection process is executed through a dedicated selection scheduler. This scheduler maintains a metadata index of all currently available reliable broadcast protocol instances and keeps it synchronized with the reputation assessment database in real time. When it is necessary to prepare input for the asynchronous Byzantine protocol consensus, the selection scheduler executes the corresponding selection strategy according to the current stage of the system (initialization or stable operation) and passes the list of selected instance identifiers to the asynchronous Byzantine protocol consensus engine.

[0120] The efficiency of the entire selection process is ensured through a multi-level caching mechanism. Popular instance data and consensus node reputation information are cached in memory to ensure fast response even under peak load. At the same time, the system adopts a distributed lock mechanism to ensure data consistency in a multi-consensus node environment and prevent selection errors caused by concurrent access.

[0121] Step 4: After processing the reliable broadcast protocol and the asynchronous Byzantine protocol in parallel, the consensus-reached cryptographic proposals are decrypted with thresholds, their correctness is verified, and they are timestamped and sorted. They are then submitted to the distributed ledger. At the same time, after each round of consensus is completed, the overall reputation value of the consensus node is adjusted through a dynamic reward and punishment mechanism based on the behavior of the consensus node in this round of consensus.

[0122] The specific process of threshold decryption of encrypted proposals is as follows:

[0123] Once the system reaches consensus on a batch of cryptographic proposals via the asynchronous Byzantine Fault Tolerance protocol, each consensus node begins collecting decrypted fragments from other consensus nodes; when any consensus node has collected no less than After obtaining a valid decrypted fragment, the plaintext transaction is reconstructed using these fragments; redundancy is eliminated and logical correctness is verified for these decrypted transactions; all transactions that pass the verification are globally sorted according to their timestamps, packaged into formal blocks, and finally submitted to the distributed ledger for persistent storage.

[0124] In a specific implementation, the technical implementation of the consensus post-processing stage includes three key aspects: transaction processing flow, dynamic reputation adjustment mechanism, and coordinated operation of parallel execution. These aspects together constitute a complete closed loop of system operation.

[0125] In the transaction processing phase, once the asynchronous Byzantine protocol reaches a consensus on a batch of cryptographic proposals, the system immediately initiates a threshold decryption process. Each consensus node begins exchanging the decrypted fragments it holds. This process employs digital signature technology to verify the authenticity and integrity of each decrypted fragment. When any consensus node collects no less than... After a valid decryption fragment is obtained, the system will reconstruct the complete symmetric key through the Lagrange interpolation algorithm, and then decrypt the encrypted transactions in batches. This threshold mechanism ensures that even if some consensus nodes maliciously withhold decryption fragments, as long as the fragments provided by honest consensus nodes reach the threshold, the decryption process can be completed smoothly.

[0126] Decrypted transactions undergo rigorous multi-layered verification. First, a redundancy check is performed, using a Bloom filter to quickly identify and eliminate duplicate transactions. Next, logical correctness checks are executed, including verifying the validity of digital signatures, checking the standardization of transaction formats, confirming the adequacy of account balances, and preventing double-spending. Verified transactions are then globally sorted according to their internal timestamps. This sorting process takes into account network transmission latency and clock synchronization errors, using an improved timestamp algorithm to ensure a consistent sequential view across the entire network. Finally, these ordered transactions are packaged into formal blocks, with Merkle tree structures used to ensure data integrity, and then submitted to the distributed ledger for persistent storage.

[0127] Furthermore, the dynamic adjustment of the overall credit score is achieved through a dynamic reward and punishment mechanism, specifically:

[0128] After each round of consensus, the behavior of the consensus nodes is immediately verified. If a consensus node provides a valid and reliable broadcast protocol instance in this round of consensus, a preset reward value is added to its original comprehensive reputation value. If a consensus node provides an invalid and reliable broadcast protocol instance, a preset penalty value is subtracted from its original comprehensive reputation value.

[0129] In a specific implementation, the core of the dynamic reward and punishment mechanism lies in achieving real-time feedback on the behavior of consensus nodes through quantified reward and punishment operations. The reward and punishment values ​​are not fixed, and their specific design is as follows:

[0130] The reward value and penalty value are defined as a fixed ratio of the consensus node's current overall reputation value. This is an adaptive design that makes the consensus node with the higher overall reputation value have a greater reputation change due to a single action, thereby accelerating the rise of high-quality consensus nodes and eliminating malicious consensus nodes.

[0131] Regarding the reward value, the increase in the overall reputation value of a consensus node when it provides a valid and reliable broadcast protocol instance is: ,in, This is the overall reputation value of the consensus node before adjustment. The reward coefficient is a preset positive decimal; the penalty value is the reduction in the overall reputation value of a consensus node when it provides an invalid instance of a reliable broadcast protocol. ,in, This is the overall reputation value of the consensus node before adjustment. The penalty coefficient is a pre-defined positive decimal.

[0132] Reward coefficient The value is set between 0.01 and 0.1, which is 1% to 10% of the current overall reputation score. This range ensures that good behavior is given significant and positive incentives. If the value is too small (e.g., less than 0.01), the incentive effect is weak, and the consensus nodes lack the motivation to improve their reputation; if... If the value is too large (e.g., greater than 0.1), it may cause drastic fluctuations in the reputation system, and a few consensus nodes may quickly form a monopoly due to their initial advantage.

[0133] Penalty coefficient The value range is set between 0.05 and 0.2, which is 5% to 20% of the current overall credit score. In practice, to ensure a deterrent effect against malicious behavior, the penalty coefficient is... It should be greater than or equal to the reward coefficient. This scope, along with harsher penalties for malicious acts ( This reflects the system's security design principles. Stronger penalties ensure that even a single or small number of malicious actions are sufficient to significantly lower a consensus node's reputation ranking, thereby quickly excluding it from the core consensus group and effectively protecting system security. Insufficient penalties will not effectively suppress malicious behavior. After implementing rewards and penalties, the overall reputation value of consensus nodes will be limited to an effective range. Internally, measures are in place to prevent the overall reputation score from overflowing due to accumulated rewards or penalties;

[0134] If the consensus node's overall reputation score fails to rank in the updated sort... If the name is not selected, the reliable broadcast protocol instance initiated by that consensus node will no longer be selected in subsequent consensus processes;

[0135] In terms of dynamic reputation adjustment, the system has established a sophisticated reward and punishment mechanism. After each round of consensus, the system will automatically audit the behavior of each consensus node. For consensus nodes that provide valid and reliable broadcast protocol instances, the system will add a preset reward value to their original comprehensive reputation value. The setting of this reward value takes into account the complexity and quality of the instance. Conversely, for consensus nodes that provide invalid instances, the system will implement graded penalties according to the severity of the invalid instances, ranging from slight deductions to significant downgrades. This differentiated reward and punishment design enables the system to accurately reflect the actual performance of consensus nodes.

[0136] It is particularly important to note that the system has a dynamic participation threshold. Before each round of consensus begins, the system recalculates the reputation ranking of all consensus nodes, and only those ranked above the top participate. Only consensus nodes that meet the highest performance can participate in the provision of reliable broadcast protocol instances in this round. This mechanism ensures that system resources are always tilted towards the best-performing consensus nodes, while also providing other consensus nodes with the incentive to regain the opportunity to participate by improving their performance.

[0137] The parallel execution mechanism of Region 1 and Region 2 is achieved through time overlap: when some reliable broadcast protocol instances in Region 1 have completed broadcasting, Region 2 immediately begins to process the consensus of these instances without waiting for all instances to complete broadcasting, while Region 1 continues to process the broadcasting of the remaining instances.

[0138] In terms of parallel execution coordination, the system achieves efficient collaboration between Region 1 and Region 2 through an event-driven architecture. When any instance in Region 1... Once a reliable broadcast protocol instance completes its broadcast, the system will immediately trigger Region 2 to begin processing the consensus process for these instances, without waiting for all instances to complete their broadcasts. This ready-to-process mechanism significantly reduces waiting time and improves the system's response speed. Meanwhile, Region 1 continues to process the broadcasts of the remaining instances, and newly completed instances can be added to the consensus process of Region 2 incrementally.

[0139] To achieve this parallel execution, the system employs multithreading technology and lock-free data structures to ensure that the two regions can run efficiently and concurrently without resource contention. The system also establishes a robust exception handling mechanism, which automatically triggers a retry or replacement process when an instance encounters a problem during the consensus process, ensuring the stable operation of the entire system. Through this sophisticated parallel coordination design, the system achieves near-linear performance improvement while maintaining Byzantine fault tolerance and safety.

[0140] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0141] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0142] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0143] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A blockchain asynchronous consensus method based on GRA-HB-BFT, characterized in that, The specific steps include: Step 1: A consensus node reputation evaluation model is constructed by using grey correlation analysis combined with entropy method. The information entropy of each evaluation index is calculated by using entropy method to determine its objective weight. The correlation between the evaluation index sequence of each consensus node and the ideal reference sequence is evaluated by grey correlation degree calculation. Finally, the comprehensive reputation value of each consensus node is output. Step 2: The asynchronous public subset protocol is decoupled into two parallel processing areas. Area one is a reliable broadcast protocol area, and area two is an asynchronous Byzantine protocol area. In area one, after each consensus node receives the transaction proposal submitted by the client, it performs threshold encryption processing to generate an encrypted proposal, which is broadcast as a reliable broadcast protocol instance. Step 3: In area two, based on the pre-constructed consensus node reputation evaluation model, the instances initiated by high-reputation consensus nodes are selected from the multiple reliable broadcast protocol instances broadcast in area one as input for subsequent consensus decision-making using the asynchronous Byzantine protocol. Step 4: After parallel processing of the reliable broadcast protocol and the asynchronous Byzantine protocol, the encrypted proposal reached consensus is threshold decrypted, correctness verified, and timestamp sorted, and submitted to the distributed ledger. At the same time, after each round of consensus is completed, the comprehensive reputation value of each consensus node is adjusted through a dynamic reward and punishment mechanism based on its performance in the current round of consensus.

2. The blockchain asynchronous consensus method based on GRA-HB-BFT according to claim 1, characterized in that: The construction of the node reputation evaluation model specifically includes: The response time, transaction processing success rate, and historical consensus participation rate of the consensus node are selected as evaluation indexes. The response time refers to the time taken from receiving a transaction request by a consensus node to completing processing and broadcasting a response to the network. The transaction processing success rate refers to the ratio of the number of transactions successfully processed and submitted to the ledger to the total number of transactions received by the consensus node in a unit of time. The historical consensus participation rate refers to the proportion of the consensus node actively participating in and completing the consensus process in the past N rounds of consensus, where N is a pre-set historical round threshold. The behavior data of each consensus node in the historical rounds is collected to form a dataset to be evaluated. The dataset to be evaluated is standardized to eliminate the dimensional differences between different evaluation indexes. Based on the standardized data, the information entropy of each evaluation index is calculated using the entropy method. According to the information entropy, the objective weight of each evaluation index is determined using the entropy method. The optimal value of each evaluation index in all consensus nodes is selected to form an ideal reference sequence. The optimal value is determined according to the evaluation index, with the maximum value for positive evaluation indexes and the minimum value for negative evaluation indexes. The grey correlation coefficient between each evaluation index value of each consensus node and the corresponding index of the ideal reference sequence is calculated. The comprehensive reputation value of each consensus node is calculated by combining the grey correlation coefficient and the objective weight of each evaluation index.

3. The blockchain asynchronous consensus method based on GRA-HB-BFT according to claim 2, characterized in that: The data standardization process uses the range standardization method, which is specifically: For each evaluation index, the maximum value and the minimum value of the evaluation index in the data set to be evaluated of all consensus nodes are identified, and the data to be evaluated of each consensus node under the evaluation index is scaled to an interval according to the maximum value and the minimum value. The information entropy of each evaluation index is calculated using the entropy method. The process is as follows: Based on the standardized values, the proportion of the standardized value of each consensus node in each evaluation index to the sum of all standardized values of the evaluation index is calculated. According to the proportion distribution, the information entropy of each evaluation index is calculated using information entropy theory. The objective weight of each evaluation index is determined by the following process: according to the information entropy value of each evaluation index, the information utility value is calculated, and the information utility value of each evaluation index is compared with the sum of the information utility values of all evaluation indexes to determine the objective weight of each evaluation index.

4. The blockchain asynchronous consensus method based on GRA-HB-BFT according to claim 3, characterized in that: The optimal value of each evaluation index is selected to constitute an ideal reference sequence wherein, is the optimal value of the ideal reference sequence on the i th evaluation index; for the i th consensus node, the standardized data sequence is wherein is the standard value of the j th evaluation index in the i th consensus node, and j is the index of the evaluation index.​ The grey correlation coefficient of the ith consensus node and the ideal reference sequence on the ith evaluation index is calculated as follows: The grey correlation coefficient of the ith consensus node and the ideal reference sequence on the ith evaluation index is calculated as follows: ； in, Let $\frac{j}{i}$ be the absolute deviation between the standard value and the ideal value of the $j$-th evaluation metric in the $i$-th consensus node. and These represent the minimum and maximum absolute deviations of the standard and ideal values ​​for all consensus nodes across all evaluation metrics, respectively. The resolution coefficient is used to adjust the resolution of the gray relational coefficient. For the first The consensus node is at the _th ... Grey relational coefficients on each evaluation indicator; The first consensus node is determined according to the comprehensive reputation value of each consensus node The following formula is used: ； wherein, is the objective weight of the i-th evaluation index, is the subjective weight of the i-th evaluation index, for subsequent ranking and selection of consensus nodes.

5. The blockchain asynchronous consensus method based on GRA-HB-BFT according to claim 4, characterized in that: The instance initiated by the consensus node with high reputation is preferably used as input, which adopts a two-stage strategy: The system is run in advance The round consensus is defined as an initialization phase, in which a random strategy is used to select A reliable broadcast protocol instance participates in the asynchronous Byzantine protocol consensus, wherein The preset round threshold is The maximum fault-tolerant consensus node number of the system is After completing the round consensus, the system enters a stable running phase, in which, according to the latest comprehensive reputation values of the consensus nodes, the reliable broadcast protocol instances published by the consensus nodes with the top comprehensive reputation values are selected to participate in the asynchronous Byzantine protocol consensus; The node reputation evaluation model performs a comprehensive recalculation every round of consensus, wherein the preset update period; between two comprehensive updates, the comprehensive reputation value of the consensus node is adjusted in a small range through a dynamic reward and punishment mechanism.

6. The blockchain asynchronous consensus method based on GRA-HB-BFT according to claim 5, characterized in that: When processing the transaction proposal in the reliable broadcast protocol area, the consensus node performs the following operations: A predetermined number of transaction proposals are randomly extracted from the locally maintained transaction pool, and threshold encryption algorithm is used to encrypt the transaction proposals; the ciphertext formed after encryption is used as the input content of the reliable broadcast protocol instance of the consensus node, and is broadcast to all other consensus nodes in the network; The specific process of threshold decryption of the encrypted proposal is as follows: When the system reaches consensus on a batch of encrypted proposals through the asynchronous Byzantine protocol, each consensus node begins to collect decryption fragments from other consensus nodes; when any consensus node collects not less than valid decryption fragments, it reconstructs the plaintext transactions using these fragments; redundancy elimination and logical correctness verification are performed on these decrypted transactions. All transactions that pass the verification are globally sorted according to their timestamps and packaged into official blocks, which are finally submitted to the distributed ledger for persistent storage.

7. The blockchain asynchronous consensus method based on GRA-HB-BFT according to claim 6, characterized in that: The dynamic adjustment of the comprehensive reputation value is realized through a dynamic reward and punishment mechanism, which is specifically as follows: After each round of consensus, the behavior of the consensus node is immediately verified; if the consensus node provides an effective reliable broadcast protocol instance in this round of consensus, a predetermined reward value is added to its original comprehensive reputation value; if the consensus node provides an invalid reliable broadcast protocol instance, a predetermined penalty value is subtracted from its original comprehensive reputation value; If the consensus node's integrated reputation value fails to enter the top of the updated ranking, the reliable broadcast protocol instance initiated by the consensus node will no longer be selected in the subsequent consensus. The parallel execution mechanism of area one and area two is realized by time overlap: when some reliable broadcast protocol instances in area one complete broadcasting, area two immediately starts processing the consensus of these instances without waiting for all instances to complete broadcasting, while area one continues to process the broadcasting of the remaining instances.

Images