An artificial intelligence-based blockchain dynamic trust evaluation system and method

Abstract

The application provides a blockchain dynamic trust evaluation system and method based on artificial intelligence, which performs time series modeling on transaction data of a blockchain node and consensus history records of the node participating in consensus, obtains behavior credibility of the blockchain node, and then calculates resource contribution degrees of each transaction node, fuses the behavior credibility and each resource contribution degree, and obtains consensus scores of each transaction node; constructs a node graph neural network of the blockchain based on a transaction topology relationship and an interaction frequency between each transaction node, detects a plurality of malicious nodes in the blockchain using the node graph neural network, and then evaluates a consensus influence coefficient of current malicious node aggregation on consensus of the blockchain nodes; and performs a punitive correction on the consensus scores of each malicious node through the consensus influence coefficient, and obtains final trust scores of each transaction node in the blockchain. Based on the above scheme, punitive correction of malicious node aggregation in blockchain node trust evaluation can be realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of blockchains, and more particularly, to a blockchain dynamic trust evaluation system and method based on artificial intelligence. BACKGROUND

[0002] The blockchain dynamic trust management utilizes the distributed ledger, non-tamperability and consensus mechanism of the blockchain, dynamically evaluates and adjusts the trust relationship between each participating subject, and when the transaction node behaves well in the blockchain network, the trust value is improved; if there is a violation, the trust value is reduced, thereby effectively encouraging the node to comply with the rules, ensuring the safe and stable operation of the blockchain network, and being suitable for various scenarios such as digital currency transactions, supply chain traceability, and Internet of Things device management.

[0003] The single reputation scoring mechanism used in the traditional blockchain trust evaluation method only calculates a static reputation value based on the independent historical behavior data of the node, ignoring the topological association characteristics and dynamic interaction mode between nodes. Therefore, it cannot identify the group attack threat formed by malicious nodes through high-frequency transaction pairs, dense interaction networks or collaborative behavior, and malicious nodes can disguise reputation through periodic compliance behavior and gather to launch collaborative attacks at critical times. Due to the lack of graph neural network analysis of transaction topological structure, dynamic modeling of time series behavior mode and quantitative evaluation of group influence, the traditional method cannot effectively detect abnormal association patterns between nodes, nor can it implement precise punishment and inhibition of collaborative attacks, ultimately leading to serious security risks for the blockchain network. Therefore, how to realize the punishment correction of malicious node gathering in blockchain node trust evaluation has become a difficult problem in the industry. SUMMARY

[0004] The present application provides a blockchain dynamic trust evaluation system and method based on artificial intelligence, which can realize the punishment correction of malicious node gathering in blockchain node trust evaluation.

[0005] In a first aspect, the present application provides a blockchain dynamic trust evaluation method based on artificial intelligence, comprising:

[0006] obtaining transaction data of a blockchain node and consensus history records of the node participating in consensus;

[0007] performing time series modeling on the transaction data and the consensus history records to obtain behavior credibility of the blockchain node, calculating resource contribution degrees of each transaction node based on verification block information of the blockchain ledger records, and fusing the behavior credibility and each resource contribution degree to obtain consensus scores of each transaction node;

[0008] A node graph neural network of a blockchain is constructed based on a transaction topological relationship and an interaction frequency between each transaction node, a plurality of malicious nodes in the blockchain are detected using the node graph neural network, and a consensus influence coefficient of a consensus of the blockchain nodes by a current malicious node aggregation is evaluated;

[0009] A consensus score of each malicious node is punished and corrected by the consensus influence coefficient, and a final trust score of each transaction node in the blockchain is obtained.

[0010] In some embodiments, the transaction data and the consensus history record are time-series modeled to obtain a behavior credibility of a blockchain node, which specifically includes:

[0011] The transaction data and the consensus history record are aligned in multiple time windows to obtain a transaction field set and a consensus field set of each time window;

[0012] An abnormal transaction proportion of each time window is determined through each transaction field set, and a consensus stability index of each time window is determined through each consensus field set;

[0013] A time-series feature matrix of the blockchain node is constructed according to all the abnormal transaction proportions and the consensus stability indexes;

[0014] The time-series feature matrix is input into a pre-trained long short-term memory network model to output the behavior credibility of the blockchain node.

[0015] In some embodiments, the resource contribution degree of each transaction node is calculated based on block information of a blockchain ledger record verification, which specifically includes:

[0016] Block information of a blockchain ledger record verification is obtained;

[0017] For each transaction node, a computing power contribution, a token pledge contribution, and a storage space contribution of the transaction node are extracted from the block information of the blockchain ledger record verification;

[0018] The resource contribution degree of the transaction node is determined through the computing power contribution, the token pledge contribution, and the storage space contribution, and the resource contribution degree of each transaction node is obtained.

[0019] In some embodiments, the consensus score of each transaction node is obtained by fusing the behavior credibility and the resource contribution degree, which specifically includes:

[0020] For each transaction node, the behavior credibility and the resource contribution degree of the transaction node are normalized;

[0021] A fusion weight of the dynamic trust score is set through a network state of the transaction node;

[0022] The normalized behavior credibility and the resource contribution degree are fused by weighting based on the fusion weight to obtain a consensus score of the transaction node, and then consensus scores of each transaction node are obtained.

[0023] In some embodiments, constructing a node graph neural network of the blockchain based on the transaction topological relationship and the interaction frequency between each transaction node specifically comprises:

[0024] Obtaining the transaction topological relationship and the interaction frequency between each transaction node;

[0025] Extracting node addresses of each transaction node from the transaction topological relationship;

[0026] Generating a node graph neural network of the blockchain by taking each node address as a graph node and each interaction frequency as an edge.

[0027] In some embodiments, detecting a plurality of malicious nodes in the blockchain by using the node graph neural network specifically comprises:

[0028] Aggregating neighbor information of each graph node in the node graph neural network;

[0029] Obtaining node embedding values in each neighbor information;

[0030] Abnormal screening of the blockchain nodes based on each node embedding value to obtain a plurality of malicious nodes.

[0031] In some embodiments, the consensus score of each malicious node is punished and corrected by the consensus influence coefficient to obtain a final trust score of each transaction node in the blockchain specifically comprises:

[0032] For each malicious node, obtaining a maximum value of all historical consensus influence coefficients of the malicious node in a historical time period as a correction reference value;

[0033] Determining a penalty coefficient of the trust score of the malicious node according to the consensus influence coefficient and the correction reference value;

[0034] Determining the trust score of the malicious node by the penalty coefficient and the consensus score of the malicious node, and then obtaining the trust score of each malicious node;

[0035] Combining the trust score of each malicious node and the consensus score of each transaction node to obtain a final trust score of each transaction node in the blockchain.

[0036] In a second aspect, the present application provides an artificial intelligence-based dynamic trust evaluation system for a blockchain, which comprises an evaluation unit, and the evaluation unit comprises:

[0037] An acquisition module is configured to acquire transaction data of a blockchain node and a consensus history record of the node participating in consensus;

[0038] A processing module is configured to perform time series modeling on the transaction data and the consensus history record to obtain behavior credibility of the blockchain node, calculate resource contribution degrees of each transaction node based on verified block information of a blockchain ledger record, and fuse the behavior credibility and the resource contribution degrees to obtain a consensus score of each transaction node;

[0039] The processing module is further configured to construct a node graph neural network of the blockchain based on a transaction topology relationship and an interaction frequency between each transaction node, detect a plurality of malicious nodes in the blockchain using the node graph neural network, and further evaluate a consensus influence coefficient of current malicious node aggregation on consensus of the blockchain nodes;

[0040] An execution module is configured to perform a punitive correction on the consensus score of each malicious node through the consensus influence coefficient to obtain a final trust score of each transaction node in the blockchain.

[0041] In a third aspect, the present application provides a computer device, which comprises a memory and a processor, the memory is configured to store a computer program, and the processor is configured to call and run the computer program from the memory, so that the computer device executes the above-mentioned artificial intelligence-based dynamic trust evaluation method of the blockchain.

[0042] In a fourth aspect, the present application provides a computer readable storage medium, which stores instructions or codes, when the instructions or codes are run on a computer, the computer executes the above-mentioned artificial intelligence-based dynamic trust evaluation method of the blockchain.

[0043] The technical scheme provided by the embodiments of the present application has the following beneficial effects:

[0044] The application provides a blockchain dynamic trust evaluation system and method based on artificial intelligence, wherein transaction data of a blockchain node and consensus history records of the node participating in consensus are acquired; the transaction data and the consensus history records are subjected to time sequence modeling to obtain behavior credibility of the blockchain node; resource contribution degrees of each transaction node are calculated based on verification of block information of a blockchain ledger record; the behavior credibility and each resource contribution degree are fused to obtain a consensus score of each transaction node; a node graph neural network of the blockchain is constructed based on a transaction topology relationship and an interaction frequency between each transaction node; the node graph neural network is used to detect a plurality of malicious nodes in the blockchain, and then a consensus influence coefficient of current malicious node aggregation on consensus of the blockchain nodes is evaluated; the consensus scores of each malicious node are subjected to a punitive correction through the consensus influence coefficient to obtain final trust scores of each transaction node in the blockchain.

[0045] Therefore, in the application, the consensus scores of each malicious node are subjected to a punitive correction through the consensus influence coefficient to obtain final trust scores of each transaction node in the blockchain; first, the consensus score can obtain a comprehensive evaluation index fusing the behavior credibility and the resource contribution degree, thereby providing a precise judgment basis for the punitive correction of malicious node aggregation, which can avoid one-sidedness of single-dimensional evaluation, for example, is vulnerable to witch attacks only relying on resources, and can identify potential malicious nodes with high resources but low credibility, thereby ensuring that the weight reduction of malicious nodes considers both individual threats and actual resource occupation of the network, and from the root, the risk that malicious nodes obtain trust through resource stacking and then launch attacks is inhibited; then, the consensus influence coefficient can obtain a quantitative threat value of current malicious node aggregation on the consensus network, thereby realizing precise attack on collaborative attacks, the consensus influence coefficient reflects individual malicious degree and captures group synergy effect, thereby effectively identifying hidden threats such as identity networks and collusion of staking pools in the blockchain in malicious attacks, and the punitive correction based on the consensus influence coefficient can dynamically adjust the weight reduction degree and apply exponential punishment to malicious groups with high density aggregation to dissolve their collaborative advantages; linear punishment is adopted for isolated malicious nodes to avoid excessive inhibition, thereby significantly improving the defense capability of the blockchain against new collaborative attacks; in summary, the above scheme can realize the punitive correction of malicious node aggregation in the blockchain node trust evaluation. BRIEF DESCRIPTION OF DRAWINGS

[0046] In order to more clearly illustrate the technical solutions in the embodiments of the application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description only constitute some embodiments of the application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0047] Figure 1 is an exemplary flowchart of an artificial intelligence-based blockchain dynamic trust evaluation method according to some embodiments of the present application;

[0048] Figure 2 is a flowchart of implementing a punitive correction according to some embodiments of the present application;

[0049] Figure 3 is a structural diagram of an evaluation unit according to some embodiments of the present application;

[0050] Figure 4 is a structural diagram of a computer device for implementing an artificial intelligence-based blockchain dynamic trust evaluation method according to some embodiments of the present application. DETAILED DESCRIPTION

[0051] In order to better understand the technical solutions of the present application, the technical solutions of the present application will be described in detail below in conjunction with the drawings in the specification and specific embodiments.

[0052] Reference Figure 1 The figure is an exemplary flowchart of an artificial intelligence-based blockchain dynamic trust evaluation method according to some embodiments of the present application, which mainly includes the following steps:

[0053] In step 101, transaction data of a blockchain node and a consensus history record of the node participating in consensus are obtained.

[0054] It should be noted that in the present application, the transaction data refers to the operation information of the transfer, the smart contract call in the blockchain network, and the transaction data contains the sender, the receiver, the amount, the timestamp, and the transaction data is the basic unit of the change of the blockchain state; the consensus history record refers to the behavior log of the transaction node in the consensus mechanism, which includes the block verification, the voting signature, the proposal response process, and reflects the contribution of the node to the network security; the transaction data of the blockchain node and the consensus history record of the node participating in the consensus are obtained from the blockchain distributed ledger.

[0055] In a specific implementation, first, a data query request is initiated through an application program interface or a full node client of a target node; if the node supports lightweight query, index information such as transaction hash and block height can be directly obtained, and then complete transaction content is pulled according to the index; if historical data archiving is required, block files need to be traversed or rely on an off-chain database. For a private chain, local ledgers can be directly accessed through permission control; for a public chain, a remote procedure call or a third-party blockchain browser interface needs to be relied on to filter target data through transaction hash or address, so as to obtain the query result as transaction data of the blockchain node; then, if the blockchain adopts a transparent consensus mechanism, the node public key and block right can be associated by analyzing the verifier signature or random number in the block header; for a consortium chain, the node proposal voting record needs to be extracted from the ordering service log. If the network provides a consensus audit interface, the node participation round and state synchronization time can be directly queried; otherwise, local log files or network broadcast messages need to be parsed, and the consensus participation time sequence is reconstructed in combination with the timestamp and block height. Some chains store evidence of malicious behavior, and the integrity of the historical record needs to be verified additionally, so as to obtain the verified historical record as the consensus history record of the blockchain node participating in the consensus.

[0056] In step 102, the transaction data and the consensus history record are time-series modeled to obtain the behavior credibility of the blockchain node, the resource contribution degree of each transaction node is calculated based on the verified block information of the blockchain ledger record, and the consensus score of each transaction node is obtained by fusing the behavior credibility and the resource contribution degree of each node.

[0057] In some embodiments, the behavior credibility of the blockchain node obtained by time-series modeling of the transaction data and the consensus history record can be achieved by the following steps:

[0058] The transaction data and the consensus history record are aligned in multiple time windows to obtain a transaction field set and a consensus field set of each time window;

[0059] The abnormal transaction proportion of each time window is determined through each transaction field set, and the consensus stability index of each time window is determined through each consensus field set;

[0060] A time-series feature matrix of the blockchain node is constructed according to all the abnormal transaction proportions and consensus stability indexes;

[0061] The time-series feature matrix is input into a pre-trained long short-term memory network model to output the behavior credibility of the blockchain node.

[0062] It should be noted that in the present application, the behavior credibility is a scalar value that comprehensively represents the credibility of the historical behavior of the node, and the higher the behavior credibility, the more reliable it represents; the transaction field set refers to a transaction data set divided according to a fixed time window; the consensus field set refers to a behavior data set of the node participating in consensus within the same time window; the abnormal transaction proportion refers to the percentage of the total number of transactions that are marked as abnormal within a single time window; and the consensus stability index is a weighted comprehensive score for quantifying the reliability of the consensus behavior of the node.

[0063] In a specific implementation, first, the transaction data and the consensus history record are aligned in multiple time windows to obtain the transaction field set and the consensus field set of each time window. This can be achieved in the following manner: a sliding time window (by default, a time window of 30s length) is set, and for each time window, the records of the time window are extracted from the transaction data and the consensus history record, respectively. After the transaction data is aggregated by time window, the statistical features of the transaction volume, the number of unique addresses, and the average transaction fee are generated as the transaction field set of the time window; the consensus data extracts the block number, the number of verification failures, the message broadcast delay, and other indicators in each time window as the consensus field set of the time window; if the data timestamps are not continuous, linear interpolation or forward filling is used to fill in the missing values to ensure that the lengths of the windows are consistent. Time zone differences and conflicting data caused by chain bifurcation need to be handled in the alignment process, and finally a structured multi-dimensional time sequence is output. The transaction field set and the consensus field set of each time window can be obtained in the above manner; second, the abnormal transaction proportion of each time window is determined through each transaction field set, and the consensus stability indicator of each time window is determined through each consensus field set. This can be achieved in the following manner: based on a preset matching rule (for example, the transaction fee deviates from the mean value by more than a specified value), each transaction field set is matched to identify abnormal transactions, and the proportion of abnormal transactions in each time window is counted as the abnormal transaction proportion of the corresponding time window, so that the abnormal transaction proportion of each time window is obtained; for each time window, the key indicators (for example, the number of consecutive blocks not mined, the voting divergence rate) of the time window are extracted from the consensus field sequence, the weights of each key indicator are determined based on historical experience, and finally a weighted average method is used to output a normalized score of 0~1 as the consensus stability indicator of the time window, so that the consensus stability indicator of each time window is obtained in the above manner; then, a time sequence feature matrix of the blockchain node is constructed according to all the abnormal transaction proportions and the consensus stability indicators. This can be achieved in the following manner: the abnormal transaction proportions and the consensus stability indicators of each time window are concatenated in time sequence to form a matrix, and each row represents a feature vector of a complete window, so that the time sequence feature matrix of the blockchain node is obtained; finally, the time sequence feature matrix is input into a pre-trained long short-term memory (LSTM) model to output the behavior credibility of the blockchain node. This can be achieved in the following manner: the pre-trained LSTM model includes an input layer (adapted to the dimension of the feature matrix), a hidden layer (to capture long-term dependencies), and a fully connected output layer (to map to the credibility). During training, labeled data (for example, historical matrices of known malicious / normal nodes) is used to optimize the parameters with cross-entropy as the loss function, and in the inference stage, the time sequence feature matrix of the target node is input into the model, the LSTM automatically learns the dynamic correlation between transaction anomalies and consensus stability (for example, an increase in abnormal transactions leads to an increase in consensus failure rate), and finally a probabilistic credibility is output as the behavior credibility of the blockchain node.

[0064] In some embodiments, the calculation of the resource contribution of each transaction node based on the verification block information recorded in the blockchain ledger can be implemented by the following steps:

[0065] Obtaining the verification block information recorded in the blockchain ledger;

[0066] For each transaction node, extracting the computing power contribution, token staking contribution and storage space contribution of the transaction node from the verification block information;

[0067] Determining the resource contribution of the transaction node by the computing power contribution, the token staking contribution and the storage space contribution, and further obtaining the resource contribution of each transaction node.

[0068] It should be noted that in the present application, the resource contribution is a weighted score quantifying the support level of each transaction node for the blockchain network resources; the verification block information refers to the recorded data of each transaction node participating in block generation and verification in the blockchain network; the computing power contribution refers to the computing resources invested by the transaction node in the consensus process; the token staking contribution refers to the number and duration of tokens locked by the transaction node in the proof-of-stake mechanism; and the storage space contribution refers to the distributed storage capacity provided by the transaction node for the blockchain network.

[0069] In a specific implementation, first, the verification block information of the blockchain ledger record can be implemented in the following manner, that is, querying the block header information in the target time range through the local ledger database of the blockchain node, extracting the fields related to verification, which includes the effective calculation amount in a unit of time, the number and duration of token locking in the proof of stake mechanism, the storage data amount and availability time. For a consortium chain, the verification record needs to be obtained from the ordering service log or smart contract event; for a public chain, the block data can be pulled in batches through the blockchain browser interface. If the data authenticity needs to be verified, multiple node ledger copies need to be compared to ensure that the information has not been tampered with, and possible conflict blocks caused by chain forks are handled, so that the relevant field results obtained by querying are used as the verification block information of the blockchain ledger record. Then, for each transaction node, the calculation power contribution, token pledge contribution and storage space contribution of the transaction node from the verification block information can be implemented in the following manner, that is, for each transaction node, the effective calculation amount of the transaction node in a unit of time in the verification block information is used as the calculation power contribution of the transaction node; the product of the number and duration of token locking of the transaction node in the proof of stake mechanism in the verification block information is used as the token pledge contribution of the transaction node; and the product of the storage data amount and availability time of the transaction node in the verification block information is used as the storage space contribution of the transaction node. Finally, the resource contribution degree of the transaction node is determined through the calculation power contribution, the token pledge contribution and the storage space contribution, and the resource contribution degree of each transaction node can be obtained in the following manner, that is, the influence weight of the calculation power contribution, the token pledge contribution and the storage space contribution on the transaction node can be preset through historical experience, so as to calculate the weighted sum of the normalized calculation power contribution, the token pledge contribution and the storage space contribution as the resource contribution degree of the transaction node, and the resource contribution degree of each transaction node can be obtained through the above manner.

[0070] In some embodiments, fusing the behavior credibility and the resource contribution degree of each transaction node to obtain the consensus score of each transaction node can be implemented in the following steps:

[0071] For each transaction node, the behavior credibility and the resource contribution degree of the transaction node are normalized;

[0072] The fusion weight of the dynamic trust score is set through the network state of the transaction node;

[0073] The normalized behavior credibility and the resource contribution degree are weighted and fused based on the fusion weight to obtain the consensus score of the transaction node, and the consensus score of each transaction node is obtained.

[0074] It should be noted that in the present application, the consensus score is an evaluation value for measuring the priority of a node participating in consensus, and the higher the consensus score, the more suitable the transaction node is to be selected as a block producer or a verifier; in specific implementation, for each transaction node, the behavior credibility and the resource contribution degree of the transaction node can be normalized by the following implementation, that is, by using the minimum-maximum normalization, the behavior credibility (such as the 0-1 probability value output by the LSTM) is directly retained, and the resource contribution degree is scaled to the same interval according to the maximum value and the minimum value of the index in all transaction nodes, so that the normalization of the behavior credibility and the resource contribution degree is completed; then, the fusion weight of the dynamic trust score can be realized by setting the network state of the transaction node in the following way, that is, the weight corresponding to the network state of the transaction node is selected from the network-weight mapping table of the blockchain as the fusion weight of the dynamic trust score, and the fusion weight is used to adjust the proportion of the behavior credibility and the resource contribution degree in the final consensus score; finally, the normalized behavior credibility and the resource contribution degree are weighted and fused based on the fusion weight to obtain the consensus score of the transaction node, and the consensus score of each transaction node can be obtained by using the following method, that is, the weighted sum of the normalized behavior credibility and the resource contribution degree is calculated as the consensus score of the transaction node using the fusion weight, and the consensus score of each transaction node can be obtained by using the above method.

[0075] In step 103, a node graph neural network of the blockchain is constructed based on the transaction topological relationship and the interaction frequency between each transaction node, and the node graph neural network is used to detect a plurality of malicious nodes in the blockchain, and then the consensus influence coefficient of the current malicious node aggregation on the consensus of the blockchain nodes is evaluated.

[0076] In some embodiments, the node graph neural network of the blockchain can be constructed based on the transaction topological relationship and the interaction frequency between each transaction node by the following steps:

[0077] Obtain the transaction topological relationship and the interaction frequency between each transaction node;

[0078] Extract the node address of each transaction node from the transaction topological relationship;

[0079] Generate the node graph neural network of the blockchain by taking each node address as a graph node and each interaction frequency as an edge.

[0080] It should be noted that in the present application, the node graph neural network is a graph structure data model taking a blockchain address as a node and an interaction frequency as an edge weight, which can be used to analyze node behavior and network characteristics through deep learning; the transaction topology relationship represents the connection structure formed between different nodes in the blockchain network through transactions; the interaction frequency refers to the message passing frequency between transaction nodes within a specified time, which can be used to quantify the association strength between transaction nodes; the node address refers to a string that uniquely identifies a participant in the blockchain, which can be used to represent an entity node in the graph neural network.

[0081] In a specific implementation, first, the transaction topology relationship and the interaction frequency between the transaction nodes can be obtained in the following manner: extracting raw transaction data from the blockchain ledger, including the sender address, receiver address, timestamp, and transaction type of each transaction, for large-scale data, using a distributed computing framework (such as the Spark framework) for batch processing, counting the number of transactions between each transaction node according to a time window (default is every hour) to form an interaction frequency matrix, if a smart contract is involved, further parsing the contract event log to identify the calling relationship between the contract and the user node as the transaction topology relationship, where the node represents the address and the edge represents the interaction frequency; then, the node address of each transaction node can be obtained from the transaction topology relationship in the following manner: taking the nodes in the transaction topology relationship as the node address of the corresponding transaction node, thereby obtaining the node address of each transaction node; finally, generating the node graph neural network of the blockchain by taking each node address as a graph node and each interaction frequency as an edge.

[0082] In some embodiments, detecting a plurality of malicious nodes in the blockchain using the node graph neural network can be achieved in the following steps:

[0083] Aggregating the neighbor information of each graph node in the node graph neural network;

[0084] Obtaining the node embedding value in each neighbor information;

[0085] Abnormal screening of the blockchain nodes based on the node embedding values to obtain a plurality of malicious nodes.

[0086] It should be noted that in the present application, the malicious node refers to a behaviorally abnormal node that threatens the security of the blockchain; the neighbor information refers to the other nodes and their interaction relationship data directly connected to the target node in the graph neural network; the node embedding value is a low-dimensional dense vector used to represent the topological structure, behavior pattern, and semantic information of the node in the global network.

[0087] In a specific implementation, first, the neighbor information of each graph node in the node graph neural network can be implemented in the following manner: for each graph node in the node graph neural network, the nodes directly connected to the graph node are selected from the graph neural network as neighbor nodes, the account activity, transaction pattern code and weight of the connection edge (i.e. the normalized transaction frequency) of each neighbor node are collected, and the mean of all normalized transaction frequencies is taken as the neighbor information of the graph node, so that the neighbor information of each graph node in the node graph neural network can be obtained; then, the node embedding value in each neighbor information can be implemented in the following manner: the mean of the normalized transaction frequency in each neighbor information is taken as the node embedding value in the corresponding neighbor information, so that the node embedding value in each neighbor information can be obtained; finally, the malicious nodes can be obtained in the following manner: the abnormal threshold of the blockchain node can be preset through historical experience, the graph node corresponding to the neighbor information with a node embedding value lower than the abnormal threshold is taken as a malicious node, and the malicious nodes can be obtained.

[0088] In some embodiments, the evaluation of the consensus influence coefficient of the current malicious node aggregation on the consensus of the blockchain node can be implemented in the following steps:

[0089] obtaining the consensus score of each malicious node;

[0090] determining the inner edge density of each malicious node in the node graph neural network;

[0091] determining the consensus influence coefficient of the current malicious node aggregation on the consensus of the blockchain node through all the consensus scores and inner edge densities.

[0092] It should be noted that in the present application, the consensus influence coefficient is a threat quantitative value that comprehensively quantifies the individual ability of the malicious node and the degree of group cooperation. In specific implementation, first, the consensus score of each malicious node can be obtained by the following method, that is, taking the consensus score of the transaction node corresponding to each malicious node as the consensus score of the malicious node, and then determining the inner edge density of each malicious node in the node graph neural network by the following method, that is, for each malicious node, taking the node graph neural network containing only malicious nodes and the edges between them as the induced subgraph of the malicious node, counting the actual interaction edge number in the induced subgraph, and then dividing by the theoretical maximum edge number (i.e. the number of edges when the malicious nodes are fully connected), so as to take the calculation result as the inner edge density of the malicious node, and through the above method, the inner edge density of each malicious node in the node graph neural network can be obtained. The inner edge density is an index value for quantifying the closeness between malicious nodes. The higher the inner edge density, the greater the possibility of collaborative maliciousness. Finally, the consensus influence coefficient of the current malicious node cluster on the consensus of the blockchain node can be determined by the sum of all consensus scores and the inner edge density by the following method, that is, taking the product of the sum of all consensus scores and the inner edge density as the consensus influence coefficient of the current malicious node cluster on the consensus of the blockchain node.

[0093] In step 104, the consensus score of each malicious node is punished and corrected by the consensus influence coefficient, and the final trust score of each transaction node in the blockchain is obtained.

[0094] In some embodiments, the consensus score of each malicious node is punished and corrected by the consensus influence coefficient, and the final trust score of each transaction node in the blockchain is obtained, as shown in Figure 2 The figure is a flowchart of the implementation of the punitive correction in some embodiments of the present application. The punitive correction in the present embodiment can be implemented by the following steps:

[0095] In step 1041, for each malicious node, the maximum value of all historical consensus influence coefficients of the malicious node in the historical time period is obtained as the correction reference value;

[0096] In step 1042, a penalty coefficient of the trust score of the malicious node is determined according to the consensus influence coefficient and the correction reference value;

[0097] In step 1043, the trust score of the malicious node is determined by the penalty coefficient and the consensus score of the malicious node, and then the trust score of each malicious node is obtained;

[0098] In step 1044, the trust score of each malicious node and the consensus score of each transaction node are combined to obtain the final trust score of each transaction node in the blockchain.

[0099] It should be noted that in the present application, the final trust score is the credibility set of the blockchain network node evaluated by the unified standard; the correction reference value is the quantitative value of the maximum potential threat degree of the malicious node to the blockchain network; the punishment coefficient is an adjustment factor for quantifying the reduction proportion of the trust score of the malicious node; and the trust score is a comprehensive score reflecting the credibility of the transaction node at the current time.

[0100] In a specific implementation, first, for each malicious node, the maximum value of all historical consensus influence coefficients of the malicious node in a historical time period is obtained as the correction reference value, which can be implemented in the following manner: for each malicious node, the maximum value of all historical consensus influence coefficients of the malicious node in a historical time period (by default, the last 1 week) is obtained as the correction reference value; second, the punishment coefficient of the trust score of the malicious node is determined according to the consensus influence coefficient and the correction reference value, which can be implemented in the following manner: the ratio of the current consensus influence coefficient to the correction reference value is taken as the punishment coefficient of the trust score of the malicious node; then, the trust score of the malicious node is determined by the punishment coefficient and the consensus score of the malicious node, and the trust score of each malicious node is obtained in the following manner: the product of the punishment coefficient and the consensus score of the malicious node is taken as the trust score of the malicious node, and the trust score of each malicious node is obtained through the above manner; finally, the trust scores of each malicious node and the consensus scores of each transaction node are combined to obtain the final trust scores of each transaction node in the blockchain, which can be implemented in the following manner: for each transaction node in the blockchain, if the transaction node is not a malicious node, the consensus score of the transaction node is directly taken as the final trust score of the transaction node, and if the transaction node is a malicious node, the trust score of the malicious node corresponding to the transaction node is taken as the final trust score of the transaction node, and the final trust scores of each transaction node in the blockchain are obtained through the above manner.

[0101] In addition, another aspect of the present application, in some embodiments, the present application provides an artificial intelligence-based dynamic trust evaluation system for a blockchain, which comprises an evaluation unit, which refers to Figure 3 The figure is a structural schematic diagram of an evaluation unit according to some embodiments of the present application, which comprises an acquisition module 201, a processing module 202 and an execution module 203, which are described as follows:

[0102] The acquisition module 201 is mainly used for acquiring the transaction data of the blockchain node and the consensus history record of the node participating in the consensus in the present application;

[0103] The processing module 202 is configured to perform time series modeling on the transaction data and the consensus history record to obtain behavior credibility of the blockchain node, calculate resource contribution degrees of each transaction node based on verified block information of the blockchain ledger record, and fuse the behavior credibility and the resource contribution degrees to obtain a consensus score of each transaction node.

[0104] It should be noted that the processing module 202 is further configured to construct a node graph neural network of the blockchain based on a transaction topology relationship and an interaction frequency between each transaction node, detect a plurality of malicious nodes in the blockchain using the node graph neural network, and further evaluate a consensus influence coefficient of a consensus of the blockchain node caused by current malicious node aggregation.

[0105] The execution module 203 is configured to perform a punitive correction on the consensus score of each malicious node based on the consensus influence coefficient to obtain a final trust score of each transaction node in the blockchain.

[0106] The above describes an example of the blockchain dynamic trust evaluation system and method based on artificial intelligence provided by the embodiments of the present application in detail. It can be understood that the corresponding device includes a hardware structure and / or a software module corresponding to the execution of each function in order to achieve the above functions. Those skilled in the art should easily realize that the units and algorithm steps of each example described in combination with the embodiments disclosed in the present application can be realized in the form of hardware or a combination of hardware and computer software. Whether a certain function is executed in the form of hardware or computer software driven hardware depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0107] In some embodiments, the present application further provides a computer device including a memory and a processor, the memory is configured to store a computer program, and the processor is configured to call and run the computer program from the memory, so that the computer device executes the above-mentioned blockchain dynamic trust evaluation method based on artificial intelligence.

[0108] In some embodiments, with reference to Figure 4 The dashed line in the figure indicates that the unit or the module is optional. The figure is a structural schematic diagram of a computer device for implementing the blockchain dynamic trust evaluation method based on artificial intelligence according to the embodiments of the present application. The blockchain dynamic trust evaluation method based on artificial intelligence described in the above embodiments can be implemented by the computer device shown in the figure. Figure 4The computer device shown is implemented by a computer device including at least one processor 301, a memory 302, and at least one communication unit 305, and can be a terminal device or a server or a chip.

[0109] The processor 301 can be a general-purpose processor or a special-purpose processor. For example, the processor 301 can be a central processing unit (CPU), which can be used to control the computer device, execute a software program, and process data of the software program. The computer device can further include a communication unit 305 to implement input (reception) and output (transmission) of signals.

[0110] For example, the computer device can be a chip, and the communication unit 305 can be an input and / or output circuit of the chip, or the communication unit 305 can be a communication interface of the chip. The chip can be a component of a terminal device or a network device or other device.

[0111] For another example, the computer device can be a terminal device or a server, and the communication unit 305 can be a transceiver of the terminal device or the server, or the communication unit 305 can be a transceiver circuit of the terminal device or the server.

[0112] One or more memories 302 can be included in the computer device, and programs 304 can be stored on the memories 302. The programs 304 can be run by the processor 301 to generate instructions 303, so that the processor 301 executes the methods described in the above method embodiments according to the instructions 303. Alternatively, data (such as a target audit model) can also be stored in the memory 302. Alternatively, the processor 301 can also read data stored in the memory 302. The data can be stored in the same storage address as the programs 304, or the data can be stored in different storage addresses from the programs 304.

[0113] The processor 301 and the memory 302 can be separately arranged or integrated together, for example, integrated on a system on chip (SOC) of the terminal device.

[0114] It should be understood that each step of the above method embodiments can be accomplished by logic circuits in the form of hardware or instructions in the form of software in the processor 301, which can be a CPU, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof.

[0115] Those skilled in the art will appreciate that embodiments of the application can be devised for a method, a system, or a computer program product. Accordingly, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage devices, etc.) embodying computer-readable program code.

[0116] For example, in some embodiments, the present application also provides a computer-readable storage medium, which stores instructions or codes, when the instructions or codes are run on a computer, cause the computer to perform the above-mentioned artificial intelligence-based blockchain dynamic trust evaluation method.

[0117] While the preferred embodiments of the application have been described, additional variations and modifications can be made to the embodiments by those skilled in the art once they learn of the basic inventive concepts. Therefore, the appended claims are intended to encompass within their scope all such variations and modifications as are included within the scope of the application.

[0118] Obviously, various modifications and changes can be made to the present application by those skilled in the art without departing from the spirit and scope of the present application. Accordingly, the present application intends to include all such modifications and changes as fall within the scope of the claims and their equivalents.

Claims

1. An artificial intelligence-based blockchain dynamic trust assessment method, characterized in that, The method comprises the following steps: obtaining transaction data of a blockchain node and consensus history records of the node participating in consensus; performing time series modeling on the transaction data and the consensus history records to obtain behavior credibility of the blockchain node, calculating resource contribution degrees of each transaction node based on verified block information of blockchain ledger records, and fusing the behavior credibility and the resource contribution degrees to obtain consensus scores of each transaction node; constructing a node graph neural network of the blockchain based on transaction topological relations and interaction frequencies between each transaction node, using the node graph neural network to detect a plurality of malicious nodes in the blockchain, and then evaluating a consensus influence coefficient of current malicious node aggregation on the consensus of the blockchain nodes; performing a punitive correction on the consensus scores of each malicious node through the consensus influence coefficient to obtain final trust scores of each transaction node in the blockchain; wherein the punitive correction on the consensus scores of each malicious node through the consensus influence coefficient to obtain the final trust scores of each transaction node in the blockchain specifically comprises: for each malicious node, obtaining a maximum value of all historical consensus influence coefficients of the malicious node in a historical time period as a correction reference value; determining a penalty coefficient of the trust score of the malicious node according to the consensus influence coefficient and the correction reference value; determining the trust score of the malicious node through the penalty coefficient and the consensus score of the malicious node, and then obtaining the trust score of each malicious node; merging the trust scores of each malicious node and the consensus scores of each transaction node to obtain the final trust scores of each transaction node in the blockchain.

2. The method of claim 1, wherein, The time series modeling on the transaction data and the consensus history records to obtain the behavior credibility of the blockchain node specifically comprises: aligning the transaction data and the consensus history records in multiple time windows to obtain transaction field sets and consensus field sets of each time window; determining abnormal transaction proportions of each time window through each transaction field set and determining consensus stability indicators of each time window through each consensus field set; constructing a time series feature matrix of the blockchain node according to all abnormal transaction proportions and consensus stability indicators; inputting the time series feature matrix into a pre-trained long short-term memory network model to output the behavior credibility of the blockchain node.

3. The method of claim 1, wherein, The calculation of the resource contribution degrees of each transaction node based on the verified block information of the blockchain ledger records specifically comprises: obtaining the verified block information of the blockchain ledger records; for each transaction node, extracting computing power contribution, token staking contribution and storage space contribution of the transaction node from the verified block information; determining the resource contribution degree of the transaction node through the computing power contribution, the token staking contribution and the storage space contribution, and then obtaining the resource contribution degrees of each transaction node.

4. The method of claim 1, wherein, The fusion of the behavior credibility and the resource contribution degrees to obtain the consensus scores of each transaction node specifically comprises: for each transaction node, normalizing the behavior credibility and the resource contribution degree of the transaction node; setting a fusion weight of the dynamic trust score through the network state of the transaction node; The normalized behavior credibility and the resource contribution degree are fused based on the fusion weight to obtain a consensus score of the transaction node, and then consensus scores of all transaction nodes are obtained.

5. The method of claim 1, wherein, The node graph neural network of the blockchain is constructed based on the transaction topological relationship and the interaction frequency between the transaction nodes, and specifically includes: Obtaining the transaction topological relationship and the interaction frequency between the transaction nodes; Extracting node addresses of the transaction nodes from the transaction topological relationship; Taking each node address as a graph node and each interaction frequency as an edge, a node graph neural network of the blockchain is generated.

6. The method of claim 1, wherein, Detecting the malicious nodes in the blockchain using the node graph neural network specifically includes: Aggregating neighbor information of each graph node in the node graph neural network; Obtaining node embedding values in each neighbor information; Based on the node embedding values, the blockchain nodes are abnormally screened to obtain the malicious nodes.

7. An artificial intelligence based blockchain dynamic trust assessment system, which employs the method of any one of claims 1 to 6 for blockchain dynamic trust assessment, the system comprising an assessment unit, characterized in that, The evaluation unit includes: An obtaining module configured to obtain transaction data of the blockchain nodes and consensus history records of the nodes participating in consensus; A processing module configured to perform time series modeling on the transaction data and the consensus history records to obtain behavior credibility of the blockchain nodes, calculate resource contribution degrees of the transaction nodes based on verified block information of the blockchain ledger records, fuse the behavior credibility and the resource contribution degrees to obtain consensus scores of the transaction nodes, and construct a node graph neural network of the blockchain based on the transaction topological relationship and the interaction frequency between the transaction nodes, detect malicious nodes in the blockchain using the node graph neural network, and further evaluate a consensus influence coefficient of the current malicious node aggregation on the consensus of the blockchain nodes; An executing module configured to punishively correct the consensus scores of the malicious nodes based on the consensus influence coefficient to obtain final trust scores of the transaction nodes in the blockchain. The computer device includes a memory and a processor, the memory is configured to store a computer program, and the processor is configured to call and run the computer program from the memory, so that the computer device executes the artificial intelligence-based dynamic trust evaluation method of the blockchain in any one of claims 1 to 6.

8. A computer device, comprising: The computer readable storage medium stores instructions or codes, and when the instructions or codes are run on the computer, the computer executes the artificial intelligence-based dynamic trust evaluation method of the blockchain in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, ​

Images