Block chain transaction semantic detection incentive mechanism based on evolutionary game

Abstract

The invention relates to the technical field of software communication, in particular to a block chain transaction semantic detection incentive mechanism based on evolutionary game, which comprises a block chain memory pool transaction semantic detection mechanism, a block chain and user decision evolutionary game model and a block chain memory pool transaction semantic detection mechanism. Comprising the following steps: adding a transaction semantic detection list between a block chain memory pool queuing list and a suspended list; the method has the advantages that beforehand detection of transactions before uplink is realized; the method tries to explore how to drive willing to actively perceive unknown risks before uplink for the first time, and designs a corresponding incentive mechanism to drive most willing to execute semantic detection before transaction uplink through an action decision of game theory modeling detection; through the evolutionary game framework, a block chain developer can predict the evolutionary condition of a user decision in a block chain network in advance, so that the developer can carry out parameter adjustment in advance according to a game evolutionary result.

Description

Technical Field

[0001] The present invention relates to the field of software communication technology, and in particular to a blockchain transaction semantic detection incentive mechanism based on evolutionary game. Background Art

[0002] Blockchain technology is a distributed ledger technology that can record and manage various digital assets, such as cryptocurrencies, NFTs (Non-Fungible Tokens), etc. It uses cryptographic algorithms and distributed storage to maintain the security, integrity and immutability of data, and realizes information sharing and exchange through P2P network protocols. The large-scale Web 3.0 digital economy represented by DeFi (Decentralized finance) and NFTs (Non-Fungible Tokens) usually relies on blockchain to realize value transfer and complex and diverse custom transaction logic. However, transaction risks or illegal behaviors such as financial fraud, hacker attacks, and money laundering are rampant in the current blockchain ecosystem, and have become the main form of financial network security confrontation between countries. According to the 2024 Chainalysis Crypto Crime Report, a total of at least US$24.2 billion in illegal transactions were found in 2023. In order to prevent systemic financial risks, my country's regulatory authorities have gradually tightened financial supervision. The People's Bank of China issued the "Financial Technology Development Plan (2022-2025)", pointing out that it is necessary to establish an intelligent risk control mechanism and implement penetrating supervision on financial technology innovation. With the complexity and diversification of blockchain transactions, understanding the intentions behind the transactions is very important for the development of the Web 3.0 blockchain ecosystem.

[0003] The extraction of complex transaction semantics and risk perception in blockchain is the premise for coping with systemic financial risks. How to fight against advanced threats is the key and difficult problem faced by the blockchain field. In the blockchain transaction system, the rapid emergence of new transactions and smart contract logics has brought huge challenges to the existing boundary detection system that relies on rules and features. The existing blockchain transaction semantic detection is a lagging supervision of transaction data that has been uploaded to the chain, and it is impossible to achieve full-process penetrating supervision. For example, Yanmei Z et al. combined bytecode features, user transactions and operation code frequencies to obtain more comprehensive features, and used the improved LightGBM algorithm to identify smart contract Ponzi schemes. In order to detect money laundering in Bitcoin mixing services, M.Mazhar Rathore et al. searched for unique features of money laundering through Bitcoin transaction graphs and developed a high-precision detection model based on C4.5 decision trees. Zhiying W et al. proposed a transaction semantic representation method based on Motif, which can capture transaction semantic information in transaction data streams in real time, effectively distinguish different transaction semantics, and can be used for various downstream tasks. However, due to the technical characteristics of blockchain that cannot be tampered with, transaction data that has been uploaded to the chain cannot be withdrawn, resulting in regulatory lags and even hindering the development of the Web 3.0 ecosystem. Therefore, it is in this context that the present invention considers the problem of pre-detection incentives for semantic risk perception of blockchain transactions. For the first time, it attempts to explore how to drive nodes to be willing to actively perceive unknown risks before uploading transactions to the chain, and models node action decisions through game theory, and designs corresponding incentive mechanisms to drive most nodes to be willing to perform semantic detection before transactions are uploaded to the chain, thereby achieving penetrating supervision of blockchain transactions. Summary of the invention

[0004] In order to solve the above technical problems, the technical solution provided by the present invention is: a blockchain transaction semantic detection incentive mechanism based on evolutionary game, including a blockchain memory pool transaction semantic detection mechanism, a blockchain node and a user decision evolutionary game model, characterized in that the blockchain memory pool transaction semantic detection mechanism includes adding a transaction semantic detection list between the blockchain memory pool queue list and the suspension list, wherein the semantic detection model performs semantic detection on the transaction, enters the transaction detected as normal into the suspension list, and deletes the transaction detected as abnormal;

[0005] The blockchain node and user decision evolution game model includes: the blockchain node and user decision game model, including the setting of semantic detection rewards and penalty coefficients, and the setting of user waiting time cost and impatience;

[0006] The evolutionary game model is used to study the evolution of node and user strategies in the blockchain network over time under different parameter settings;

[0007] The complete cycle of existing Bitcoin transactions is as follows: users initiate transactions in their wallets, such as sending funds to another wallet, and sign the transaction with their private key; the signed transaction is broadcast to a node on the blockchain; the node will check and verify the transaction, and if the check passes, it will be added to the node memory pool and broadcast to peer nodes; each peer node that receives the transaction will perform the same operation, verifying and propagating the transaction on the network; some of the nodes are nodes that pack the transaction into blocks, and then compete to solve the difficult problem. Once the node successfully solves the problem, the new block will be broadcast on the network; although there is a risk of transactions being rolled back due to forks, the risk of being rolled back will become smaller and smaller as time goes by; it is generally believed that after 6 consecutive blocks are successfully on the chain after the transaction is on the chain, the transaction is confirmed.

[0008] As an improvement, the steps of the evolutionary game model are as follows:

[0009] S1. Basic assumptions

[0010] The semantic detection incentive mechanism in the present invention will reward R1 for each transaction detected by the node after successful verification. R1 is a constant value that can be set in advance. If the abnormal transaction is uploaded to the chain because the node does not perform semantic detection and causes losses to the user, the semantic detection incentive mechanism will punish the node; although different abnormal transactions cause different losses to users, the penalty coefficient γ in this incentive mechanism is a preset constant value, and the heterogeneity of abnormal transactions is not considered for the time being; in the incentive mechanism of the present invention, it is not considered that after two blocks are mined at the same time, a fork is needed to determine the longest chain, which will cause some transactions to be rolled back; the semantic detection model used in the present invention is a neural network model. Due to limitations such as data imbalance, limited data volume, and model complexity, the accuracy of the model is difficult to achieve 100%, and the model is set The probability of detecting a normal transaction as an abnormal transaction (false positive rate) is F, the probability of detecting a normal transaction as a normal transaction (true negative rate) is 1-F, the probability of detecting an abnormal transaction as an abnormal transaction (true positive rate) is T, and the probability of detecting an abnormal transaction as a normal transaction (false negative rate) is 1-T. These data can be obtained by training and verifying the transaction semantic detection model; α means the node computing power ratio, which represents the ratio of the node's hash computing power to the total computing power of the blockchain network, which is approximately equal to the probability that the node successfully solves the difficult problem first and puts the block on the chain. The closer α is to 1, the stronger the node's hash computing power is, and the easier it is to mine the block first.

[0011] S2. Game participants

[0012] The participants in the game are composed of two parties, namely the node that solves the difficult problem and the user who initiates the transaction, represented by Participant I and Participant II respectively. The former's profit comes from detection incentives and transaction fees paid by users, while the latter's profit comes from the successful on-chain launch of the initiated transaction and rewards for malicious behavior;

[0013] S3, Strategy Space

[0014] Suppose the strategy set of the node is S I = {Execute detection, do not execute detection}, where execute detection means that the node first uses the semantic detection model to perform semantic detection on transactions and filters out transactions with abnormal detection results when packaging blocks, and do not execute detection means that the node does not perform semantic detection and filtering operations on transactions when packaging blocks; the user's strategy set is S II ={initiate normal transaction, initiate abnormal transaction}, where initiating normal transaction means that the user forwards normal transaction to the node, and the transaction will not cause loss to any user of the blockchain network, and initiating abnormal transaction means that the user forwards the transaction to the node for financial fraud, extortion, money laundering and other operations; for the sake of demonstration, let A, B, C, and D represent the four strategies of the node performing detection and not performing detection before packaging the block, and the user initiating normal transaction and initiating abnormal transaction when initiating the transaction;

[0015] S4. Profit Matrix

[0016] 1) The node strategy is to execute the semantic detection model to obtain the benefits

[0017] When a node performs detection before packaging a transaction, it will first pay a detection cost, set as C1; when the user's strategy is to initiate a normal transaction, it is noted that the semantic detection model is not completely accurate. The semantic detection model has a probability of F to identify a normal transaction as an abnormal transaction; if the model detects this transaction as abnormal and deletes it, the node will not be able to get the transaction fee for the transaction, but because the node's memory pool will continue to receive new transactions, there will be new transactions immediately as a supplement, so for the node, deleting the abnormal transaction only requires a detection cost of C1; if the model accurately identifies this transaction as a normal transaction, the transaction will enter the pending list, waiting for the node to package it into the block and solve the difficult problem. The node has a probability of α to successfully package the transaction on the chain. After packaging and chaining, the node will receive the detection incentive R1 and the transaction fee f nor ; If other nodes solve the difficulty problem first, then the node will not get the corresponding detection incentive and transaction fee; use U AC It represents the utility function of the node when the node's strategy is to perform detection and the user's strategy is to initiate normal transactions. In summary, U AC =-C1+(1-F)α(R1+f nor );

[0018] When the user's strategy is to initiate an abnormal transaction, let the probability that the semantic detection model accurately identifies the abnormal transaction as an abnormal transaction be T; if the semantic detection model identifies the transaction as an abnormal transaction and deletes it, although there will be new transactions detected as normal to enter the pending list as a supplement, since the initiator of the abnormal transaction hopes that the transaction will be on the chain as soon as possible, the transaction fee will usually be set higher. Therefore, when the node deletes the abnormal transaction, it will lose the transaction fee Δf that the abnormal transaction has more than the normal transaction; if the semantic detection model identifies the transaction as a normal transaction, after the verification is successful, the node will obtain the detection incentive R1 and the transaction fee f abn , the utility function of the node is U AD =-C1+(1-T)α(R1+f abn )+TαΔf;

[0019] 2) The node strategy is to not execute the benefits of the semantic detection model

[0020] When a node does not perform detection when packaging a transaction, it will not pay the computational cost of transaction detection, nor will it receive detection incentives. When the user's strategy is to initiate a normal transaction, if the node successfully solves the difficulty problem first, it will receive the transaction fee f for this transaction. nor If the node fails to verify, it will not get the transaction fee for this transaction. At this time, the utility function of the node is U BC =αf nor ; When the user's strategy is to initiate abnormal transactions, since the node does not detect the transactions it has packaged, illegal transactions will cause losses to blockchain users after being uploaded to the chain, so this incentive mechanism considers punishing this behavior; suppose the probability that an abnormal transaction is discovered by the blockchain network after being uploaded to the chain is β, and each time a transaction in a packaged block is found to be an abnormal transaction, a penalty of γ is imposed on the node; In summary, when the node strategy is not to perform semantic detection and the user's strategy is to initiate abnormal transactions, the utility function of the node is U BD =α(f abn -βγ);

[0021] 3) User strategy is to initiate normal transaction income

[0022] When the user's strategy is to initiate a normal transaction, when the transaction is successfully uploaded to the chain, the user will obtain a successful transaction income R2; suppose the average waiting time for the transaction confirmation after the user initiates a normal transaction is t1. Since the user initiates a normal transaction, the average transaction fee cost to be paid is f nor ; At this time, if the node performs semantic detection, the semantic detection model may identify the normal transaction sent by the user as an abnormal transaction and filter it out. The user will not get the transaction success benefit R2, and will not pay the transaction fee cost accordingly, but the user needs a maximum waiting time to confirm whether the transaction has been deleted; let tmax is the maximum waiting time of the user, μ is the user's impatience coefficient, then μt max is the loss of the user's transaction after it is deleted; if the semantic detection model correctly identifies this transaction as a normal transaction, the transaction enters the queue list and is packaged by the node, and there is a probability of α that it will be successfully verified on the chain; in summary, the user's utility function V AC =(1-F)α(R2-μt1-f nor )+F(-μt max );

[0023] If the node's strategy is not to perform semantic detection, the user's transaction will not be deleted due to misclassification by the semantic detection model; since the semantic detection time is generally much shorter than the node verification time, the extended node waiting time due to the execution of semantic detection is not considered for the time being; at this time, the user's utility function is V BC =α(R2-μt1-f nor ).

[0024] 4) User strategy is to initiate abnormal transactions to gain benefits

[0025] When the user's strategy is to initiate an abnormal transaction, when the transaction is successfully uploaded to the chain, the user will receive a malicious reward R3; generally speaking, the user forwards abnormal transactions to the node because of "airdrops", extortion, scams, etc., and these purposes will bring the user far more benefits than initiating normal transactions, so R3>R2;

[0026] If the node performs semantic detection, the semantic detection model will accurately identify and delete abnormal transactions. At this time, the user will not receive malicious rewards, and will not pay transaction fees accordingly. max is the maximum waiting time cost after the user's transaction is deleted; if the semantic detection model mistakenly identifies an abnormal transaction as a normal transaction, the user's benefit is (1-T)R3; in order to confirm the transaction faster, the user usually pays a higher transaction fee. Let the transaction fee cost paid by the user who initiated the abnormal transaction be f abn =(1+θ)f nor , θ obeys the uniform distribution on [0,1]; the corresponding waiting time for transaction confirmation will also decrease with the increase of transaction fees. Let the waiting time cost be t2, and t2<t1; at this time, the user's utility function is V AD =(1-T)α(R3-μt2-f abn )+T(-μt max );

[0027] If the node does not perform semantic detection, the abnormal transactions forwarded by the user will not be detected and deleted by the semantic detection model. Once the node successfully solves the problem first, the user will receive a malicious reward; at this time, the user's utility function is V BD =α(R3-μt2-f abn ).

[0028] Compared with the existing technology, the advantages of the present invention are as follows: the present invention proposes a semantic detection mechanism for the blockchain memory pool, which realizes the pre-detection of transactions before they are put on the chain; it is the first attempt to explore how to drive nodes to be willing to actively perceive unknown risks before they are put on the chain, and through game theory modeling of the action decision of node detection, a corresponding incentive mechanism is designed to drive most nodes to be willing to perform semantic detection before transactions are put on the chain; blockchain developers can use this evolutionary game framework to predict the evolution of nodes and user decisions in the blockchain network in advance under certain parameter settings, which helps developers to adjust parameters in advance according to the game evolution results, thereby incentivizing nodes to actively perform transaction semantic detection when packaging blocks, and users are increasingly inclined to initiate normal transactions rather than illegal transactions, thereby realizing penetrating supervision of blockchain transactions, making the blockchain network more prosperous and healthy. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 This is the complete cycle diagram of the current Bitcoin transaction;

[0030] Figure 2 This is a diagram of the transaction semantic detection framework proposed by the present invention; DETAILED DESCRIPTION

[0031] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0032] The present application fully explains the problems of the prior art in the technical background, and there are no superordinate or general problems. The problems to be solved by the present application are described in detail in the following invention content and specific implementation methods, the technical problems to be solved by the technical scheme are clarified, and it is determined that the technical scheme of the present application has beneficial effects relative to the objective prior art.

[0033] See also Figure 1-Figure 2 , an embodiment provided by the present invention:

[0034] A blockchain transaction semantic detection incentive mechanism based on evolutionary game, including a blockchain memory pool transaction semantic detection mechanism, a blockchain node and a user decision evolutionary game model, characterized in that the blockchain memory pool transaction semantic detection mechanism includes adding a transaction semantic detection list between the blockchain memory pool queue list and the suspension list, wherein the semantic detection model performs semantic detection on transactions, enters transactions detected as normal into the suspension list, and deletes transactions detected as abnormal;

[0035] The blockchain node and user decision evolution game model includes: the blockchain node and user decision game model, including the setting of semantic detection rewards and penalty coefficients, and the setting of user waiting time cost and impatience;

[0036] The evolutionary game model is used to study the evolution of node and user strategies in the blockchain network over time under different parameter settings;

[0037] The complete cycle of current Bitcoin transactions is as follows: users initiate transactions in their wallets, such as sending funds to another wallet, and sign the transaction with their private key; the signed transaction is broadcast to a node on the blockchain; the node will check and verify the transaction, and if the check passes, it will be added to the node memory pool and broadcast to peer nodes; each peer node that receives the transaction will perform the same operation, verifying and propagating the transaction on the network; some of these nodes are node nodes, which pack the transaction into blocks and then compete to solve the difficult problem. Once the node successfully solves the problem, the new block will be broadcast on the network; although there is a risk of transactions being rolled back due to forks, the risk of being rolled back will become smaller and smaller as time goes by; it is generally believed that after 6 consecutive blocks are successfully on the chain after the transaction is on the chain, the transaction is confirmed.

[0038] The steps of the evolutionary game model are as follows:

[0039] The symbols used in the modeling process are shown in Table 1.

[0040]

[0041] Table 1: Explanation of symbols

[0042] S1. Basic assumptions

[0043] The semantic detection incentive mechanism in the present invention will reward the node with R1 for each transaction it detects after successful verification. R1 is a constant value that can be set in advance. If the abnormal transaction is uploaded to the chain and causes losses to the user because the node does not perform semantic detection, the semantic detection incentive mechanism will punish the node; although different abnormal transactions cause different losses to users, the penalty coefficient γ in this incentive mechanism is a preset constant value, and the heterogeneity of abnormal transactions is not considered for the time being; in the incentive mechanism of the present invention, it is not considered that after two blocks are mined at the same time, a fork is needed to determine the longest chain, which causes some transactions to be rolled back; the semantic detection model used in the present invention is a neural network model. Due to limitations such as data imbalance, limited data volume, and model complexity, the accuracy of the model is limited. It is difficult to achieve 100% accuracy. Let the probability that the model detects a normal transaction as an abnormal transaction (false positive rate) be F, the probability that the model detects a normal transaction as a normal transaction (true negative rate) be 1-F, the probability that the model detects an abnormal transaction as an abnormal transaction (true positive rate) be T, and the probability that the model detects an abnormal transaction as a normal transaction (false negative rate) be 1-T. These data can be obtained by training and verifying the transaction semantic detection model; α means the proportion of node computing power, which represents the ratio of the node's hash computing power to the total computing power of the blockchain network, which is approximately equal to the probability that the node successfully solves the difficult problem first and puts the block on the chain. The closer α is to 1, the stronger the node's hash computing power is, and the easier it is to mine the block first.

[0044] S2. Game participants

[0045] The participants in the game are composed of two parties, namely the node that solves the difficult problem and the user who initiates the transaction, represented by Participant I and Participant II respectively. The former's profit comes from detection incentives and transaction fees paid by users, while the latter's profit comes from the successful on-chain launch of the initiated transaction and rewards for malicious behavior;

[0046] S3, Strategy Space

[0047] Suppose the strategy set of the node is S I = {Execute detection, do not execute detection}, where execute detection means that the node first uses the semantic detection model to perform semantic detection on transactions and filters out transactions with abnormal detection results when packaging blocks, and do not execute detection means that the node does not perform semantic detection and filtering operations on transactions when packaging blocks; the user's strategy set is S II={initiate normal transaction, initiate abnormal transaction}, where initiating normal transaction means that the user forwards normal transaction to the node, and the transaction will not cause loss to any user of the blockchain network, and initiating abnormal transaction means that the user forwards the transaction to the node for financial fraud, extortion, money laundering and other operations; for the sake of demonstration, let A, B, C, and D represent the four strategies of the node performing detection and not performing detection before packaging the block, and the user initiating normal transaction and initiating abnormal transaction when initiating the transaction;

[0048] S4. Profit Matrix

[0049] 1) The node strategy is to execute the semantic detection model to obtain the benefits

[0050] When a node performs detection before packaging a transaction, it will first pay a detection cost, set to C1; when the user's strategy is to initiate a normal transaction, it is noted that the semantic detection model is not completely accurate. The semantic detection model has a probability of F to identify a normal transaction as an abnormal transaction; if the model detects this transaction as abnormal and deletes it, the node will not be able to obtain the transaction fee for the transaction, but because the node's memory pool will continue to receive new transactions, there will be new transactions immediately as a supplement, so for the node, deleting the abnormal transaction only requires a detection cost of C1; if the model accurately identifies this transaction as a normal transaction, the transaction will enter the pending list, waiting for the node to package it into the block and solve the difficult problem. The node has a probability of α to successfully package the transaction on the chain. After packaging and chaining, the node will receive the detection incentive R1 and the transaction fee f nor ; If other nodes solve the difficulty problem first, then the node will not get the corresponding detection incentive and transaction fee; use U AC It represents the utility function of the node when the node's strategy is to perform detection and the user's strategy is to initiate normal transactions. In summary, U AC =-C1+(1-F)α(R1+f nor );

[0051] When the user's strategy is to initiate an abnormal transaction, let the probability that the semantic detection model accurately identifies the abnormal transaction as an abnormal transaction be T; if the semantic detection model identifies the transaction as an abnormal transaction and deletes it, although there will be new transactions detected as normal to enter the pending list as a supplement, since the initiator of the abnormal transaction hopes that the transaction will be on the chain as soon as possible, the transaction fee will usually be set higher. Therefore, when the node deletes the abnormal transaction, it will lose the transaction fee Δf that the abnormal transaction has more than the normal transaction; if the semantic detection model identifies the transaction as a normal transaction, after the verification is successful, the node will obtain the detection incentive R1 and the transaction fee f abn , the utility function of the node is U AD =-C1+(1-T)α(R1+f abn )+TαΔf;

[0052] 2) The node strategy is to not execute the benefits of the semantic detection model

[0053] When a node does not perform detection when packaging a transaction, it will not pay the computational cost of transaction detection, nor will it receive detection incentives. When the user's strategy is to initiate a normal transaction, if the node successfully solves the difficulty problem first, it will receive the transaction fee f for this transaction. nor If the node fails to verify, it will not get the transaction fee for this transaction. At this time, the utility function of the node is U BC =αf nor ; When the user's strategy is to initiate abnormal transactions, since the node does not detect the transactions it packages, illegal transactions will cause losses to blockchain users after being uploaded to the chain. Therefore, this incentive mechanism considers punishing this behavior; suppose the probability that an abnormal transaction is discovered by the blockchain network after being uploaded to the chain is β. Every time a transaction in a packaged block is found to be an abnormal transaction, a penalty of β is imposed on the node; In summary, when the node strategy is not to perform semantic detection and the user's strategy is to initiate abnormal transactions, the utility function of the node is U BD =α(f abn -βγ); the profit matrix of honest nodes is shown in Table 2.

[0054]

[0055] Table 2: Payoff Matrix for Absenteeism

[0056] 3) User strategy is to initiate normal transaction income

[0057] When the user's strategy is to initiate a normal transaction, when the transaction is successfully uploaded to the chain, the user will obtain a successful transaction income R2; suppose the average waiting time for the transaction confirmation after the user initiates a normal transaction is t1. Since the user initiates a normal transaction, the average transaction fee cost to be paid is f nor ; At this time, if the node performs semantic detection, the semantic detection model may identify the normal transaction sent by the user as an abnormal transaction and filter it out. The user will not get the transaction success benefit R2, and will not pay the transaction fee cost accordingly, but the user needs a maximum waiting time to confirm whether the transaction has been deleted; let t max is the maximum waiting time of the user, μ is the user's impatience coefficient, then μt max is the loss of the user's transaction after it is deleted; if the semantic detection model correctly identifies this transaction as a normal transaction, the transaction enters the queue list and is packaged by the node, and there is a probability of α that it will be successfully verified on the chain; in summary, the user's utility function V AC =(1-F)α(R2-μt1-f nor )+F(-μt max );

[0058] If the node's strategy is not to perform semantic detection, the user's transaction will not be deleted due to misclassification by the semantic detection model; since the semantic detection time is generally much shorter than the node verification time, the extended node waiting time due to the execution of semantic detection is not considered for the time being; at this time, the user's utility function is V BC =α(R2-μt1-f nor ).

[0059] 4) User strategy is to initiate abnormal transactions to gain benefits

[0060] When the user's strategy is to initiate an abnormal transaction, when the transaction is successfully uploaded to the chain, the user will receive a malicious reward R3; generally speaking, the user forwards abnormal transactions to the node because of "airdrops", extortion, scams, etc., and these purposes will bring the user far more benefits than initiating normal transactions, so R3>R2;

[0061] If the node performs semantic detection, the semantic detection model will accurately identify and delete abnormal transactions. At this time, the user will not receive malicious rewards, and will not pay transaction fees accordingly. max is the maximum waiting time cost after the user's transaction is deleted; if the semantic detection model mistakenly identifies an abnormal transaction as a normal transaction, the user's benefit is (1-T)R3; in order to confirm the transaction faster, the user usually pays a higher transaction fee. Let the transaction fee cost paid by the user who initiated the abnormal transaction be f abn =(1+θ)f nor , θ obeys the uniform distribution on [0,1]; the corresponding waiting time for transaction confirmation will also decrease with the increase of transaction fees. Let the waiting time cost be t2, and t2<t1; at this time, the user's utility function is V AD =(1-T)α(R3-μt2-f abn )+T(-μt max );

[0062] If the node does not perform semantic detection, the abnormal transactions forwarded by the user will not be detected and deleted by the semantic detection model. Once the node successfully solves the problem first, the user will receive a malicious reward; at this time, the user's utility function is V BD =α(R3-μt2-f abn ).

[0063] The user's profit matrix is ​​shown in Table 3.

[0064]

[0065] Table 3: User benefit matrix

[0066] Evolutionary Game Modeling

[0067] Let x (0≤x≤1) represent the proportion of nodes that perform detection, 1-x represent the proportion of nodes that do not perform detection, y (0≤y≤1) represent the proportion of users who initiate normal transactions, and 1-y represent the proportion of users who initiate abnormal transactions. For the convenience of representation, let ΔR = R3-R2, Δt = t2-t1.

[0068] When nodes take the test, their average benefit (fitness) is U IA =uU AC +(1-y)U AD When nodes do not perform detection, their average benefit (fitness) is U IB =yU BC +(1-y)U BD The average benefit (average fitness) of a node is Φ I =xU IA +(1-x)U IB According to the Malthusian equation, the replication dynamic equation of the node performing the detection is taken as

[0069]

[0070] When users initiate normal transactions, their average revenue is V IIC =xV AC +(1-x)V BC When users initiate abnormal transactions, their average profit is V IID =xV AD +(1-x)V BD The average revenue of a user is Φ II =yV IIC +(1-y)V IID According to the Malthusian equation, the replication dynamic equation of the dishonest node that does not perform detection is

[0071] According to the above two replication dynamic equations, the evolutionary stable strategy (ESS) of the evolutionary game between nodes and users can be solved. This evolutionary game model constitutes the theoretical framework of this semantic detection incentive mechanism. By setting different detection incentives R1 and penalty coefficients γ, blockchain developers can clearly predict the strategic evolution of nodes executing detection or not executing detection, and users initiating normal transactions or abnormal transactions in the blockchain network.

[0072] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the application disclosed herein. The present application is intended to cover any variations, uses or adaptations of the present application, which follow the general principles of the present application and include common knowledge or customary techniques in the art that are not disclosed in the present application. The specification and examples are intended to be exemplary only, and the true scope of the present application is indicated by the claims.

[0073] It should be understood that the present application is not limited to the precise structure described above and shown in the drawings, and various modifications and changes can be made without departing from its scope. The above-described embodiments of the present application do not constitute a limitation on the scope of protection of the present application.

Claims

1. A blockchain transaction semantic detection incentive mechanism based on evolutionary game, including a blockchain memory pool transaction semantic detection mechanism, a blockchain node and a user decision evolutionary game model, characterized in that: The blockchain memory pool transaction semantic detection mechanism includes adding a transaction semantic detection list between the blockchain memory pool queue list and the suspension list, wherein the semantic detection model performs semantic detection on the transaction, enters the transaction detected as normal into the suspension list, and deletes the transaction detected as abnormal; The blockchain node and user decision evolution game model includes: the blockchain node and user decision game model, including the setting of semantic detection rewards and penalty coefficients, and the setting of user waiting time cost and impatience; The evolutionary game model is used to study the evolution of miner nodes and user strategies in the blockchain network over time under different parameter settings; The complete cycle of existing Bitcoin transactions is as follows: users initiate transactions in their wallets, such as sending funds to another wallet, and sign the transaction with their private key; the signed transaction is broadcast to a node on the blockchain; the node will check and verify the transaction, and if the check passes, it will be added to the node memory pool and broadcast to peer nodes; each peer node that receives the transaction will perform the same operation, verifying and propagating the transaction on the network; some of the nodes are nodes that pack the transaction into blocks, and then compete to solve the difficult problem. Once the node successfully solves the problem, the new block will be broadcast on the network; although there is a risk of transactions being rolled back due to forks, the risk of being rolled back will become smaller and smaller as time goes by; it is generally believed that after 6 consecutive blocks are successfully on the chain after the transaction is on the chain, the transaction is confirmed.

2. According to claim 1, a blockchain transaction semantic detection incentive mechanism based on evolutionary game, characterized in that: The steps of the evolutionary game model are as follows: S1. Basic assumptions The semantic detection incentive mechanism in the present invention will reward the node with R1 for each transaction it detects after successful verification. R1 is a constant value that can be set in advance. If the abnormal transaction is uploaded to the chain and causes losses to the user because the node does not perform semantic detection, the semantic detection incentive mechanism will punish the node; although different abnormal transactions cause different losses to users, the penalty coefficient γ in this incentive mechanism is a preset constant value, and the heterogeneity of abnormal transactions is not considered for the time being; in the incentive mechanism of the present invention, it is not considered that after two blocks are mined at the same time, a fork is needed to determine the longest chain, which causes some transactions to be rolled back; the semantic detection model used in the present invention is a neural network model. Due to limitations such as data imbalance, limited data volume, and model complexity, the accuracy of the model is limited. It is difficult to achieve 100% accuracy. Let the probability that the model detects a normal transaction as an abnormal transaction (false positive rate) be F, the probability that the model detects a normal transaction as a normal transaction (true negative rate) be 1-F, the probability that the model detects an abnormal transaction as an abnormal transaction (true positive rate) be T, and the probability that the model detects an abnormal transaction as a normal transaction (false negative rate) be 1-T. These data can be obtained by training and verifying the transaction semantic detection model; α means the proportion of node computing power, which represents the ratio of the node's hash computing power to the total computing power of the blockchain network, which is approximately equal to the probability that the node successfully solves the difficult problem first and puts the block on the chain. The closer α is to 1, the stronger the node's hash computing power is, and the easier it is to mine the block first. S2. Game participants The participants in the game are composed of two parties, namely the node that solves the difficult problem and the user who initiates the transaction, represented by Participant I and Participant II respectively. The former's profit comes from detection incentives and transaction fees paid by users, while the latter's profit comes from the successful on-chain launch of the initiated transaction and rewards for malicious behavior; S3, Strategy Space Suppose the strategy set of the node is S I = {Execute detection, do not execute detection}, where execute detection means that the node first uses the semantic detection model to perform semantic detection on transactions and filters out transactions with abnormal detection results when packaging blocks, and do not execute detection means that semantic detection and filtering operations are not performed on transactions when packaging blocks; the user's strategy set is S II ={initiate normal transaction, initiate abnormal transaction}, where initiating normal transaction means that the user forwards normal transaction to the node, and the transaction will not cause loss to any user of the blockchain network, and initiating abnormal transaction means that the user forwards the transaction to the node for financial fraud, extortion, money laundering and other operations; for the sake of demonstration, let A, B, C, and D represent the four strategies of the node performing detection and not performing detection before packaging the block, and the user initiating normal transaction and initiating abnormal transaction when initiating the transaction; S4. Profit Matrix 1) The node strategy is to execute the semantic detection model to obtain the benefits When a node performs detection before packaging a transaction, it will first pay a detection cost, set to C1; when the user's strategy is to initiate a normal transaction, it is noted that the semantic detection model is not completely accurate. The semantic detection model has a probability of F to identify a normal transaction as an abnormal transaction; if the model detects this transaction as abnormal and deletes it, it will not get the transaction fee for the transaction, but because the node's memory pool will continue to receive new transactions, there will be new transactions immediately as a supplement, so for the node, deleting the abnormal transaction only requires a detection cost of C1; if the model accurately identifies this transaction as a normal transaction, the transaction will enter the pending list, waiting to be packaged into the block and solve the difficulty problem. The node has a probability of α to successfully package the transaction on the chain. After packaging and chaining, the node will receive the detection incentive R1 and the transaction fee f nor ; If other nodes solve the difficulty problem first, then the node will not get the corresponding detection incentive and transaction fee; use U AC It represents the utility function of the node when the node's strategy is to perform detection and the user's strategy is to initiate normal transactions. In summary, U AC =-C1+(1-F)α(R1+f nor ); When the user's strategy is to initiate an abnormal transaction, let the probability that the semantic detection model accurately identifies an abnormal transaction as an abnormal transaction be T; If the semantic detection model identifies the transaction as an abnormal transaction and deletes it, although there will be new transactions detected as normal to enter the pending list as a supplement, since the initiator of the abnormal transaction hopes that the transaction will be on the chain as soon as possible, the transaction fee will usually be set higher. Therefore, when the node deletes the abnormal transaction, it will lose the transaction fee Δf that is more than the normal transaction; if the semantic detection model identifies the transaction as a normal transaction, after the verification is successful, the node will receive the detection incentive R1 and the transaction fee f abn , the utility function of the node is U AD =-C1+(1-T)α(R1+f abn )+TαΔf; 2) The node strategy is to not execute the benefits of the semantic detection model When a node does not perform detection when packaging a transaction, it will not pay the computational cost of transaction detection, nor will it receive detection incentives. When the user's strategy is to initiate a normal transaction, if the node successfully solves the difficulty problem first, it will receive the transaction fee f for this transaction. nor If the node fails to verify, it will not get the transaction fee for this transaction. At this time, the utility function of the node is U BC =αf nor ; When the user's strategy is to initiate abnormal transactions, since the node does not detect the transactions it has packaged, illegal transactions will cause losses to blockchain users after being uploaded to the chain, so this incentive mechanism considers punishing this behavior; suppose the probability that an abnormal transaction is discovered by the blockchain network after being uploaded to the chain is β, and each time a transaction in a packaged block is found to be an abnormal transaction, a penalty of γ is imposed on the node; In summary, when the node strategy is not to perform semantic detection and the user's strategy is to initiate abnormal transactions, the utility function of the node is U BD =α(f abn -βγ); 3) User strategy is to initiate normal transaction income When the user's strategy is to initiate a normal transaction, when the transaction is successfully uploaded to the chain, the user will obtain a successful transaction income R2; suppose the average waiting time for the transaction confirmation after the user initiates a normal transaction is t1. Since the user initiates a normal transaction, the average transaction fee cost to be paid is f nor ; At this time, if the node performs semantic detection, the semantic detection model may identify the normal transaction sent by the user as an abnormal transaction and filter it out. The user will not get the transaction success benefit R2, and will not pay the transaction fee cost accordingly, but the user needs a maximum waiting time to confirm whether the transaction has been deleted; let t max is the maximum waiting time of the user, μ is the user's impatience coefficient, then μt max is the loss of the user's transaction after it is deleted; if the semantic detection model correctly identifies this transaction as a normal transaction, the transaction enters the queue list and is packaged by the node, and there is a probability of α that it will be successfully verified on the chain; in summary, the user's utility function V AC =(1-F)α(R2-μt1-f nor )+F(-μt max ); If the node's strategy is not to perform semantic detection, the user's transaction will not be deleted due to misclassification by the semantic detection model; since the semantic detection time is generally much shorter than the node verification time, the extended node waiting time due to the execution of semantic detection is not considered for the time being; at this time, the user's utility function is V BC =α(R2-μt1-f nor ). 4) User strategy is to initiate abnormal transactions to gain benefits When the user's strategy is to initiate an abnormal transaction, when the transaction is successfully uploaded to the chain, the user will receive a malicious reward R3; generally speaking, the user forwards abnormal transactions to the node because of "airdrops", extortion, scams, etc., and these purposes will bring the user far more benefits than initiating normal transactions, so R3>R2; If the node performs semantic detection, the semantic detection model will accurately identify and delete abnormal transactions. At this time, the user will not receive malicious rewards, and will not pay transaction fees accordingly. max is the maximum waiting time cost after the user's transaction is deleted; if the semantic detection model mistakenly identifies an abnormal transaction as a normal transaction, the user's benefit is (1-F)R3; in order to confirm the transaction faster, the user usually pays a higher transaction fee. Let the transaction fee cost paid by the user who initiated the abnormal transaction be f abn =(1+θ)f nor , θ obeys the uniform distribution on [0, 1]; the corresponding waiting time for transaction confirmation will also decrease with the increase of transaction fees. Let the waiting time cost be t2, and t2<t1; at this time, the user's utility function is V AD =(1-T)α(R3-μt2-f abn )+T(-μt max ); If the node does not perform semantic detection, the abnormal transactions forwarded by the user will not be detected and deleted by the semantic detection model. Once the node successfully solves the problem first, the user will receive a malicious reward; at this time, the user's utility function is V BD =α(R3-μt2-f abn ).

Images