Block chain transaction privacy protector penetration model training method, penetration method and equipment

Abstract

The invention provides a block chain transaction privacy protector penetration model training method, a penetration method and equipment, and the training method comprises the steps: carrying out the processing of a transaction privacy protector in a block chain network for each piece of single address translation transaction data corresponding to each piece of entity data, respectively extracting each piece of corresponding directed graph feature data; and training a graph attention neural network model based on each piece of directed graph feature data corresponding to each piece of single address translation transaction data, so as to train the graph attention neural network model into a block chain transaction privacy protector penetration model used for outputting abnormal transaction recognition result data corresponding to the single address translation transaction data. According to the method and the device, the model for identifying whether the transaction data processed by the block chain transaction privacy protector is abnormal or not can be obtained through training, so that the penetration of the block chain transaction privacy protector can be realized based on the model, and the traceability and security of transaction execution of the transaction privacy protector are improved.

Description

Technical Field

[0001] The present application relates to the field of blockchain technology, and in particular to a blockchain transaction privacy protector penetration model training method, penetration method and device. Background Art

[0002] A transaction privacy protector (also known as a blockchain transaction privacy protector) refers to a tool used to provide transaction privacy protection functions to users through smart contracts in a blockchain network, allowing users to hide the source and destination of their financial transactions to avoid disclosing their specific deposit addresses or other transaction details, thereby achieving privacy protection for user financial transactions.

[0003] Specifically, the transaction processing services provided by the transaction privacy protector include: a smart contract only accepts a fixed amount of funds deposited; user A deposits funds into a smart contract and obtains a certificate R. User A can use R to withdraw the funds from the smart contract later without disclosing the specific deposit. However, if users other than A obtain certificate R, they can also obtain the right to control the funds from the transaction privacy protector. For example, user A, user B, and user C all deposit into the transaction privacy protector and obtain certificate R. At this time, user D obtains user A's certificate R. User D can obtain the funds from the smart contract corresponding to the transaction privacy protector, and the outside world does not know which funds user D has obtained. In other words, due to the untraceability feature of the transaction privacy protector, it is difficult for external observers to track the flow of funds, thereby increasing the anonymity of the transaction. The transaction privacy protector is like a black box, hiding the corresponding relationship between the inflow and outflow of funds. Because of this feature of the transaction privacy protector, it also becomes one of the obstacles to the identification of abnormal transactions in the blockchain.

[0004] Therefore, there is an urgent need to design a penetration technology for the transaction privacy protector. The penetration technology refers to a technology that can explore the relationship between the funds flowing into the transaction privacy protector and the funds flowing out of the transaction privacy protector. In other words, there is an urgent need to design a method that can identify whether the transaction data processed by the transaction privacy protector in the blockchain network is abnormal based on the above-mentioned inherent characteristics of the transaction privacy protector. Summary of the invention

[0005] In view of this, the embodiments of the present application provide a blockchain transaction privacy protector penetration model training method, penetration method and device to eliminate or improve one or more defects existing in the prior art.

[0006] One aspect of the present application provides a blockchain transaction privacy protector penetration model training method, including:

[0007] For each single address conversion transaction data corresponding to each entity data obtained after being processed by the transaction privacy protector in the blockchain network, each directed graph feature data corresponding to each single address conversion transaction data is extracted respectively; wherein the entity data includes: an address transaction set of a starting transaction address corresponding to a transaction account and an address transaction set corresponding to each other transaction address generated after the starting transaction address is converted at least once by the transaction privacy protector, and the address transaction set includes each transaction data corresponding to the transaction address uniquely corresponding to the address transaction set;

[0008] A graph attention neural network model is trained based on each directed graph feature data corresponding to each of the single address conversion transaction data, so that the graph attention neural network model is trained as a blockchain transaction privacy protector penetration model that outputs abnormal transaction identification result data corresponding to the single address conversion transaction data according to each of the directed graph feature data corresponding to the single address conversion transaction data.

[0009] In some embodiments of the present application, before extracting the directed graph feature data corresponding to each of the single address conversion transaction data corresponding to each of the entity data obtained after being processed by the transaction privacy protector in the blockchain network, the method further includes:

[0010] The real entity data corresponding to each transaction account obtained after being processed by the transaction privacy protector in the blockchain network are marked as positive samples respectively; and the new entity data generated after cross-construction of the address transaction set of the entity data corresponding to each transaction account is marked as a negative sample;

[0011] Based on the number of times each of the entity data marked as the positive sample and the negative sample is subjected to address conversion by the transaction privacy protector, each of the entity data is split respectively to obtain each single address conversion transaction data corresponding to each of the entity data; wherein the single address conversion transaction data includes: a transaction set before address conversion corresponding to the source address and a transaction set after address conversion corresponding to the target address; the source address is any transaction address in the entity data, and the target address is another transaction address generated after the source address is subjected to an address conversion by the transaction privacy protector.

[0012] In some embodiments of the present application, for each single address conversion transaction data corresponding to each entity data obtained after being processed by the transaction privacy protector in the blockchain network, extracting each directed graph feature data corresponding to each single address conversion transaction data respectively includes:

[0013] According to the transaction time sequence corresponding to each of the transaction data, based on a preset transaction quantity division threshold, the transaction set before address conversion and the transaction set after address conversion corresponding to each of the single address conversion transaction data are respectively divided into transaction data, so as to obtain directed graphs corresponding to the transaction set before address conversion and the transaction set after address conversion, wherein each of the directed graphs includes vertices for representing different transaction data and edges for representing transaction relationships between different transaction data;

[0014] The graph feature data corresponding to each of the directed graphs are extracted respectively to obtain the directed graph feature data corresponding to each of the single address conversion transaction data; wherein the directed graph feature data includes a directed graph corresponding to the corresponding single address conversion transaction data and the graph feature data corresponding to the directed graph.

[0015] In some embodiments of the present application, respectively extracting graph feature data corresponding to each of the directed graphs includes:

[0016] The vertex feature data of each vertex and the edge feature data of each edge corresponding to each of the directed graphs are extracted respectively to form the graph feature data corresponding to each of the directed graphs.

[0017] In some embodiments of the present application, the graph attention neural network model includes:

[0018] A graph attention neural network is used to extract features of the vertices and the edges in the directed graph for each directed graph feature data corresponding to the input single address conversion transaction data, so as to obtain graph readout embedding vectors corresponding to each directed graph feature data corresponding to the single address conversion transaction data;

[0019] A first concatenation layer is used to concatenate graph-read embedding vectors corresponding to each directed graph feature data corresponding to the transaction set before address conversion, so as to obtain a concatenation vector corresponding to the transaction set before address conversion;

[0020] The second concatenation layer is used to concatenate the graph read-out embedding vectors corresponding to the directed graph feature data corresponding to the post-address conversion transaction set, so as to obtain a concatenation vector corresponding to the post-address conversion transaction set;

[0021] A first multi-layer perceptron is used to extract transaction behavior features from the concatenation vector corresponding to the pre-address conversion transaction set output by the first concatenation layer, so as to obtain pre-address conversion transaction behavior feature data corresponding to the pre-address conversion transaction set;

[0022] A second multi-layer perceptron is used to extract transaction behavior features from the concatenated vector corresponding to the post-address conversion transaction set output by the second concatenated layer, so as to obtain post-address conversion transaction behavior feature data corresponding to the post-address conversion transaction set;

[0023] And, a cosine loss layer is used to calculate the cosine loss value between the transaction behavior characteristic data before address conversion output by the first multi-layer perceptron and the transaction behavior characteristic data after address conversion output by the second multi-layer perceptron based on the cosine loss function.

[0024] In some embodiments of the present application, the graph attention neural network model further includes:

[0025] A binary classifier is used to determine whether the cosine loss value output by the cosine loss layer is within a preset transaction behavior difference threshold range. If so, it outputs abnormal transaction identification result data indicating that there is no abnormality in the single address conversion transaction data; if not, it outputs abnormal transaction identification result data indicating that there is an abnormality in the single address conversion transaction data.

[0026] Another aspect of the present application provides a blockchain transaction privacy protector penetration method, comprising:

[0027] Obtaining each single address conversion transaction data corresponding to the entity data obtained after being processed by the transaction privacy protector in the blockchain network, and respectively extracting each directed graph feature data corresponding to each of the single address conversion transaction data;

[0028] The directed graph feature data corresponding to each single address conversion transaction data corresponding to the entity data are input into the blockchain transaction privacy protector penetration model, so that the blockchain transaction privacy protector penetration model outputs the abnormal transaction identification result data corresponding to each single address conversion transaction data corresponding to the entity data, wherein the blockchain transaction privacy protector penetration model is pre-trained based on the blockchain transaction privacy protector penetration model training method.

[0029] In some embodiments of the present application, the blockchain transaction privacy protector penetration method further includes:

[0030] From the abnormal transaction identification result data corresponding to each of the single address conversion transaction data corresponding to the entity data, the number of abnormal transaction identification result data indicating that the single address conversion transaction data does not have abnormality is obtained as the abnormal transaction value;

[0031] Determine whether the ratio between the abnormal transaction value corresponding to the entity data and the total number of single address conversion transaction data corresponding to the entity data is less than an abnormal transaction ratio threshold;

[0032] If the ratio is less than the abnormal transaction ratio threshold and the ratio is 0, entity recognition result data indicating that the entity data has no abnormal risk is generated;

[0033] If the ratio is less than the abnormal transaction ratio threshold and the ratio is not 0, generating entity recognition result data indicating that the abnormal risk level of the entity data is a first risk level;

[0034] If the ratio is equal to or greater than the abnormal transaction ratio threshold, entity recognition result data indicating that the abnormal risk level of the entity data is a second risk level is generated.

[0035] The third aspect of the present application provides a blockchain transaction privacy protector penetration model training device, including:

[0036] The training data processing module is used to extract the directed graph feature data corresponding to each single address conversion transaction data corresponding to each entity data obtained after being processed by the transaction privacy protector in the blockchain network; wherein the entity data includes: the address transaction set of the starting transaction address corresponding to the transaction account and the address transaction set corresponding to each other transaction address generated after the starting transaction address is converted at least once by the transaction privacy protector, and the address transaction set includes each transaction data corresponding to the transaction address uniquely corresponding to the address transaction set;

[0037] The model training module is used to train a graph attention neural network model based on each directed graph feature data corresponding to each of the single address conversion transaction data, so as to train the graph attention neural network model into a blockchain transaction privacy protector penetration model that outputs abnormal transaction identification result data corresponding to the single address conversion transaction data according to each of the directed graph feature data corresponding to the single address conversion transaction data.

[0038] The fourth aspect of the present application provides a blockchain transaction privacy protector penetration device, including:

[0039] An application data processing module is used to obtain each single address conversion transaction data corresponding to the entity data obtained after being processed by the transaction privacy protector in the blockchain network, and to extract each directed graph feature data corresponding to each of the single address conversion transaction data;

[0040] The model recognition module is used to input each directed graph feature data corresponding to each single address conversion transaction data corresponding to the entity data into the blockchain transaction privacy protector penetration model, so that the blockchain transaction privacy protector penetration model outputs the abnormal transaction identification result data corresponding to each single address conversion transaction data corresponding to the entity data, wherein the blockchain transaction privacy protector penetration model is pre-trained based on the blockchain transaction privacy protector penetration model training method.

[0041] The fifth aspect of the present application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the blockchain transaction privacy protector penetration model training method is implemented, and / or the blockchain transaction privacy protector penetration method is implemented.

[0042] The sixth aspect of the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the blockchain transaction privacy protector penetration model training method and / or implements the blockchain transaction privacy protector penetration method.

[0043] The seventh aspect of the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the blockchain transaction privacy protector penetration model training method and / or implements the blockchain transaction privacy protector penetration method.

[0044] The blockchain transaction privacy protector penetration model training method provided in the present application is to extract the directed graph feature data corresponding to each single address conversion transaction data corresponding to each entity data obtained after being processed by the transaction privacy protector in the blockchain network; wherein the entity data includes: an address transaction set of a starting transaction address corresponding to a transaction account and an address transaction set corresponding to other transaction addresses generated after the starting transaction address is converted at least once by the transaction privacy protector, and the address transaction set contains each transaction data corresponding to the transaction address uniquely corresponding to the address transaction set; a graph attention neural network model is trained based on each directed graph feature data corresponding to each single address conversion transaction data, so that the graph attention neural network model is trained as a blockchain transaction privacy protector penetration model for outputting abnormal transaction identification result data corresponding to the single address conversion transaction data according to each directed graph feature data corresponding to the single address conversion transaction data, so that a model for identifying whether there is an abnormality in the transaction data processed by the blockchain transaction privacy protector can be trained, and then the penetration of the blockchain transaction privacy protector can be realized based on the model, thereby improving the traceability and security of transactions executed by the transaction privacy protector.

[0045] Additional advantages, purposes, and features of the present application will be partially described in the following description, and will become partially apparent to those skilled in the art after studying the following, or may be learned from the practice of the present application. The purposes and other advantages of the present application can be achieved and obtained by the structures specifically pointed out in the specification and the drawings.

[0046] Those skilled in the art will understand that the purposes and advantages that can be achieved by the present application are not limited to the above specific description, and the above and other purposes that can be achieved by the present application will be more clearly understood based on the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The drawings described herein are used to provide a further understanding of the present application, constitute a part of the present application, and do not constitute a limitation of the present application. The components in the drawings are not drawn to scale, but are only for the purpose of illustrating the principles of the present application. In order to facilitate the illustration and description of some parts of the present application, the corresponding parts in the drawings may be enlarged, that is, they may become larger relative to other components in the exemplary device actually manufactured according to the present application. In the drawings:

[0048] Figure 1 This is a first flow chart of a blockchain transaction privacy protector penetration model training method in one embodiment of the present application.

[0049] Figure 2This is a second flow chart of the blockchain transaction privacy protector penetration model training method in one embodiment of the present application.

[0050] Figure 3 This is a schematic diagram of each directed graph corresponding to the transaction set before address conversion and the transaction set after address conversion in an example of the present application.

[0051] Figure 4 This is a third flow chart of the blockchain transaction privacy protector penetration model training method in one embodiment of the present application.

[0052] Figure 5 A schematic diagram of the architecture of a graph attention neural network model in one embodiment of the present application.

[0053] Figure 6 Another architectural schematic diagram of the graph attention neural network model in one embodiment of the present application.

[0054] Figure 7 This is a first flow chart of a blockchain transaction privacy protector penetration method in one embodiment of the present application.

[0055] Figure 8 This is a second flow chart of the blockchain transaction privacy protector penetration method in one embodiment of the present application.

[0056] Fig. 9 This is a first structural diagram of a blockchain transaction privacy protector penetration model training device in one embodiment of the present application.

[0057] Fig.10 This is a second structural diagram of the blockchain transaction privacy protector penetration model training device in one embodiment of the present application. DETAILED DESCRIPTION

[0058] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the implementation modes and the accompanying drawings. Here, the illustrative implementation modes and descriptions of the present application are used to explain the present application, but are not intended to limit the present application.

[0059] It should also be noted here that in order to avoid obscuring the present application due to unnecessary details, only the structures and / or processing steps closely related to the scheme according to the present application are shown in the accompanying drawings, while other details that are not very relevant to the present application are omitted.

[0060] It should be emphasized that the term “include / comprises” when used herein refers to the presence of features, elements, steps or components, but does not exclude the presence or addition of one or more other features, elements, steps or components.

[0061] It should also be noted that, unless otherwise specified, the term “connection” herein may refer not only to a direct connection but also to an indirect connection involving an intermediate.

[0062] Hereinafter, embodiments of the present application will be described with reference to the accompanying drawings. In the accompanying drawings, the same reference numerals represent the same or similar components, or the same or similar steps.

[0063] In one or more embodiments of the present application, the transaction privacy protector (i.e., blockchain transaction privacy protector) can be implemented in a blockchain network using a privacy protection contract based on zero-knowledge proof (Zero-Knowledge Proof, ZKP).

[0064] In one example, the working process of the transaction privacy protector includes:

[0065] (1) Deposit: Users deposit their funds or other tradable entity unique identifiers into the smart contract address provided by the transaction privacy protector.

[0066] (2) Generate zero-knowledge proofs: When making a deposit, users generate a “zero-knowledge proof” (zk-SNARKs), a cryptographic technique used to prove the authenticity of certain information without revealing specific details. Transaction privacy protectors use these proofs to ensure that users can provide valid credentials when withdrawing funds without revealing their transaction history.

[0067] (3) Privacy protection process: The smart contract of the transaction privacy protector protects the privacy of the source and destination of deposits, converts the transaction address of deposits, and prevents external observers from tracking the transaction flow.

[0068] (4) Withdrawal: Users can withdraw their deposits from the transaction privacy protector to their new transaction address. When withdrawing, users use the previously generated zero-knowledge proof to verify their deposits without revealing their specific deposit address or other transaction details.

[0069] The features of the transaction privacy protector include:

[0070] (1) Decentralization: The transaction privacy protector runs on Ethereum smart contracts and does not rely on centralized service providers, which means there is no single control point or operator.

[0071] (2) Privacy protection: By using zero-knowledge proof technology, the transaction privacy protector can effectively hide the source and destination of transactions and provide stronger privacy protection.

[0072] (3) Untraceability: Due to the complexity of the privacy protection process and the use of zero-knowledge proof, it is difficult for external observers to track the flow of funds, thereby increasing the anonymity of transactions.

[0073] The funds flowing into the transaction privacy protector cannot be tracked. The transaction privacy protector is like a black box, hiding the corresponding relationship between the inflow and outflow of funds. Because of this feature of the transaction privacy protector, it has also become one of the obstacles to combating abnormal trading behavior.

[0074] "Penetration" technology refers to a technology that can dig out the relationship between funds flowing into and out of a transaction privacy protector. Since the transaction privacy protector is an open source, decentralized, zero-knowledge proof-based technology, it is almost impossible to "positively penetrate" the transaction privacy protector. However, an entity (i.e., a user account) has its own behavioral habits, which may also be reflected in its use of the blockchain network. Therefore, this application is based on the "entity behavior invariance" assumption, which is that even if an entity uses a transaction privacy protector, the address behavior habits after privacy protection will not change.

[0075] Based on this, in order to design a method that can identify whether transaction data processed by a transaction privacy protector in a blockchain network is abnormal, the embodiments of the present application respectively provide a blockchain transaction privacy protector penetration model training method, a blockchain transaction privacy protector penetration method, a blockchain transaction privacy protector penetration model training device for executing the blockchain transaction privacy protector penetration model training method, a blockchain transaction privacy protector penetration device for executing the blockchain transaction privacy protector penetration method, a physical device, a computer-readable storage medium and a computer program product, and train the graph attention network using the disclosed entity data of the publicly used transaction privacy protector.

[0076] It can be understood that in one or more embodiments of the present application, blockchain transaction privacy protector penetration refers to identifying whether there are abnormal transactions in the entity data processed by the transaction privacy protector in the blockchain network.

[0077] The details are described in detail through the following examples.

[0078] Based on this, the embodiment of the present application provides a blockchain transaction privacy protector penetration model training method that can be implemented by a blockchain transaction privacy protector penetration model training device, see Figure 1 The blockchain transaction privacy protector penetration model training method specifically includes the following contents:

[0079] Step 100: For each single address conversion transaction data corresponding to each entity data obtained after processing by the transaction privacy protector in the blockchain network, extract each directed graph feature data corresponding to each single address conversion transaction data; wherein the entity data includes: an address transaction set of a starting transaction address corresponding to a transaction account and an address transaction set corresponding to other transaction addresses generated after the starting transaction address is converted at least once by the transaction privacy protector, and the address transaction set contains each transaction data corresponding to the transaction address uniquely corresponding to the address transaction set.

[0080] In one or more embodiments of the present application, the single address conversion transaction data includes: an address transaction set of a transaction address in the entity data (i.e., a transaction set before address conversion corresponding to the source address) and an address transaction set of another transaction address generated after the transaction address is converted once by the transaction privacy protector (i.e., a transaction set after address conversion corresponding to the target address).

[0081] Step 200: A graph attention neural network model is trained based on each directed graph feature data corresponding to each of the single address conversion transaction data, so as to train the graph attention neural network model into a blockchain transaction privacy protector penetration model for outputting abnormal transaction identification result data corresponding to the single address conversion transaction data according to each of the directed graph feature data corresponding to the single address conversion transaction data.

[0082] The graph attention neural network model is a machine learning model based on the graph attention neural network GAT. The core working principle of the graph attention neural network GAT is to calculate the relationship between nodes through the attention mechanism. In traditional neural networks, the state update of each vertex in a directed graph is performed independently. In the graph attention neural network GAT, the state update of each vertex will take into account the state of its neighboring vertices. The graph attention neural network GAT will calculate the attention weight between a vertex and its neighboring vertices, and then update the state of the vertex based on this weight. The way of updating information by calculating weights allows GAT to better capture the structural information in directed graphs.

[0083] From the above description, it can be seen that the blockchain transaction privacy protector penetration model training method provided in the embodiment of the present application can train a model for identifying whether there are anomalies in the transaction data processed by the blockchain transaction privacy protector, and then can realize the penetration of the blockchain transaction privacy protector based on the model, thereby improving the traceability and security of transactions executed by the transaction privacy protector.

[0084] In order to further improve the application effectiveness and reliability of single address conversion transaction data, in a blockchain transaction privacy protector penetration model training method provided in an embodiment of the present application, see Figure 2 , the blockchain transaction privacy protector penetration model training method further specifically includes the following contents before step 100:

[0085] Step 010: The real entity data corresponding to each transaction account obtained after being processed by the transaction privacy protector in the blockchain network are marked as positive samples respectively; and the new entity data generated after cross-construction of the address transaction set of the entity data corresponding to each transaction account is marked as a negative sample.

[0086] Specifically, the embodiment of the present application can pre-collect the entity data of each entity that has been publicly disclosed in the network and has used the transaction privacy protector. In order to ensure the training accuracy, the transaction address (i.e., Ethereum address, also referred to as address) used in this application must meet a certain transaction quantity TC. Therefore, each transaction address in the selected entity data corresponds to an address transaction set that contains multiple transaction data. It can be understood that the transaction data refers to data containing a certain transaction information, and the transaction information may include information such as transaction amount, average transaction interval time, account balance, fund retention rate, and smart contract call preference, which can be set according to actual application requirements.

[0087] In one example, the data structure of the entity data is as follows:

[0088] TXSET addr(1) :{Tx1,Tx2…}→TXSET addr(2) :{Tx1,Tx2…}→…

[0089] →TXSET addr(n) :{Tx1,Tx2…}

[0090] In the above structure, TXSET addr(1) :{Tx1,Tx2…} represents the address transaction set corresponding to address addr(1), which contains specific transaction data. The transaction data Tx1 and Tx2 in each address transaction set only represent the first and second transactions in the set respectively. Although Tx1 and Tx2 in different sets have the same symbol expression, they represent different transaction data.

[0091] The following symbol “→” represents the use of the transaction privacy protector. Due to the use of the transaction privacy protector, its address may change (or may not change). Here, addr(2) is used to represent the address that may change. If the address does not change, then:

[0092] addr(1)=addr(2)

[0093] The entity data may use the transaction privacy protector multiple times, and each use will generate a new arrow pointing to a new address transaction set. It can be seen that in the above structure, the transaction privacy protector is used n-1 times in total, forming n address transaction sets.

[0094] Step 020: Based on the number of times each of the entity data marked as the positive sample and the negative sample is subjected to address conversion by the transaction privacy protector, each of the entity data is split separately to obtain each single address conversion transaction data corresponding to each of the entity data; wherein the single address conversion transaction data includes: a transaction set before address conversion corresponding to the source address and a transaction set after address conversion corresponding to the target address; the source address is any transaction address in the entity data, and the target address is another transaction address generated after the source address is subjected to an address conversion by the transaction privacy protector.

[0095] It can be understood that the transaction set before address conversion includes various transaction data in which the source address participates; and the transaction set after address conversion includes various transaction data in which the target address participates.

[0096] In an example, based on the above example of entity data, the specific form of splitting it into each single address conversion transaction data is shown in Table 1:

[0097] Table 1

[0098]

[0099]

[0100] Assuming that N entity data are collected, each entity data can be split into n-1 single address conversion transaction data, then the total training data volume of this application can reach N×(n-1).

[0101] In order to further improve the application effectiveness and reliability of extracting each directed graph feature data corresponding to each of the single address conversion transaction data, in a blockchain transaction privacy protector penetration model training method provided in an embodiment of the present application, see Figure 2 , step 100 in the blockchain transaction privacy protector penetration model training method specifically includes the following contents:

[0102] Step 110: According to the transaction time sequence corresponding to each of the transaction data, based on a preset transaction quantity division threshold, the transaction data before address conversion and the transaction set after address conversion corresponding to each of the single address conversion transaction data are divided respectively to obtain directed graphs corresponding to the transaction set before address conversion and the transaction set after address conversion, wherein each of the directed graphs includes vertices for representing different transaction data and edges for representing transaction relationships between different transaction data.

[0103] Specifically, for each address transaction set in each training data, this application decomposes it based on the transaction number threshold T. That is, the packaging timestamps of all transaction data in the address transaction set are flattened on the timeline, and a subgraph (i.e., a directed graph) is generated for every T transactions.

[0104] See also Figure 3 , each piece of training data is protected by a transaction privacy protector ( Figure 3 The big black dot on the time axis of the left and right sides is used to divide the address transaction sets on the left and right sides into two time intervals. For the address transaction sets on the left and right sides, they are divided according to the transaction number division threshold T=3, that is, every 3 transactions are a division interval and a directed graph SG can be generated. i ( Figure 3 A total of 5 directed graphs are generated, numbered as SG 1 -SG 5 ).exist Figure 3 In the example, transactions in different intervals are marked on the timeline with dots of different colors. Since the address usually changes after the transaction privacy protector, Figure 3 All directed graphs on the left use red vertices to represent transaction addresses before address conversion by the transaction privacy protector. Figure 3 The directed graph on the right side of the middle uses dark green vertices to represent transaction addresses after address conversion by the transaction privacy protector.

[0105] Among them, regarding negative samples, this application randomly selects different entity data and cross-constructs negative sample data of the same amount as positive samples. For example, the address transaction set of entity data X is spliced ​​with the address transaction set of entity data Y.

[0106] Step 120: Extract the graph feature data corresponding to each of the directed graphs respectively to obtain the directed graph feature data corresponding to each of the single address conversion transaction data; wherein the directed graph feature data includes a directed graph corresponding to the corresponding single address conversion transaction data and the graph feature data corresponding to the directed graph.

[0107] In order to further extract the application effectiveness and reliability of the graph feature data corresponding to each of the directed graphs, in a blockchain transaction privacy protector penetration model training method provided in an embodiment of the present application, see Figure 4 , step 120 in the blockchain transaction privacy protector penetration model training method specifically includes the following contents:

[0108] Step 121: extracting vertex feature data of each vertex and edge feature data of each edge corresponding to each directed graph respectively, so as to construct graph feature data corresponding to each directed graph.

[0109] Specifically, for each directed graph SG i , all transaction data within 3 hops are completed for it, and features are extracted from the vertices and edges in the directed graph based on the completed directed graph.

[0110] In one embodiment of the present application, the vertex feature data includes: feature data for representing the in-degree of the vertex in the directed graph, the out-degree of the vertex in the directed graph, the total deposits and total withdrawals corresponding to the vertex, the average transaction interval time, the account balance, the fund retention rate and the smart contract call preference.

[0111] The fund retention rate refers to the ability of the address to process funds. For example, if an address transfers 5 ETH to this address, and the account immediately transfers all 4.7 ETH out, then the fund retention rate of this address is (5-4.7) / 5. If there are multiple transactions sending money to this address, the fund retention rate is the average of these retention rates.

[0112] Smart contract call preference is a vector whose dimension is the classification of smart contracts. At present, there have been a lot of studies that can classify contracts according to the code content of smart contracts. For example, this application can use a pre-trained language model to classify the content of smart contracts. This application divides the types of smart contracts into: finance, governance, identity authentication and authorization, supply chain, games, NFTs, contracts and laws, data and other contracts, so the smart contract call preference is a 9-dimensional vector. The vector content is the normalized result data after the number of times the address calls various types of smart contracts.

[0113] In one embodiment of the present application, the edge feature data includes: feature data for indicating the number of calls and the total amount corresponding to the edge. The number of calls corresponding to the edge refers to the number of edges after multiple transactions are merged. The feature data of the total amount refers to the total amount after multiple transactions are merged.

[0114] In order to further identify whether the transaction data processed by the blockchain transaction privacy protector has abnormal validity and reliability, in a blockchain transaction privacy protector penetration model training method provided in an embodiment of the present application, see Figure 5 The graph attention neural network model in the blockchain transaction privacy protector penetration model training method specifically includes the following contents:

[0115] A graph attention neural network is used to extract features of the vertices and the edges in the directed graph for each directed graph feature data corresponding to the input single address conversion transaction data, so as to obtain graph readout embedding vectors (Embedding) corresponding to each directed graph feature data corresponding to the single address conversion transaction data;

[0116] The first concatenation layer is used to concatenate graph-read embedding vectors corresponding to each directed graph feature data corresponding to the transaction set before address conversion, so as to obtain a concatenation vector (Embedding_left) corresponding to the transaction set before address conversion;

[0117] The second concatenation layer is used to concatenate the graph read-out embedding vectors corresponding to each directed graph feature data corresponding to the post-address conversion transaction set, so as to obtain the concatenation vector (Embedding_right) corresponding to the post-address conversion transaction set;

[0118] A first multi-layer perceptron is used to extract transaction behavior features from the concatenated vector corresponding to the pre-address conversion transaction set output by the first concatenation layer, so as to obtain pre-address conversion transaction behavior feature data (Behaviour_left) corresponding to the pre-address conversion transaction set;

[0119] The second multi-layer perceptron is used to extract transaction behavior features from the splicing vector corresponding to the post-address conversion transaction set output by the second splicing layer, so as to obtain post-address conversion transaction behavior feature data (Behaviour_right) corresponding to the post-address conversion transaction set; wherein both the first multi-layer perceptron and the second multi-layer perceptron adopt multi-layer perceptron MLP.

[0120] And, a cosine loss layer (Cosine Loss), which is used to calculate the cosine loss value between the transaction behavior characteristic data before the address conversion output by the first multi-layer perceptron and the transaction behavior characteristic data after the address conversion output by the second multi-layer perceptron based on the cosine loss function.

[0121] exist Figure 5In the figure, Embedding_1, Embedding_2, Embedding_3, Embedding_4 and Embedding_5 respectively represent the characteristic vectors corresponding to different directed graphs (i.e., graph read-out embedding vectors).

[0122] like Figure 5 As shown, all SG i The input graph attention neural network GAT extracts the respective graph readout embedding vectors Embedding. Then these embedding vectors are spliced ​​in chronological order before and after the use of the transaction privacy protector. In order to use the behavior data before the transaction privacy protector, this application inputs the spliced ​​embedding vectors into the multi-layer perceptron MLP. Note that there are two MLPs here, representing the behavior extraction before and after the use of the transaction privacy protector. At this time, the feature vectors output by the two MLPs are of the same dimension, and then their losses are evaluated through the cosine loss function and the entire network is updated.

[0123] In this application, the attention mechanism in the graph attention neural network GAT is also improved. Generally speaking, the traditional graph attention neural network GAT calculates the vertex V i With vertex V j The attention is to use the concatenation of the embedding vectors between two vertices without considering the edge between the two vertices. The original formula is as follows:

[0124]

[0125] in is the attention weight matrix, W is the graph embedding vector weight matrix, is the vertex eigenvector. The result of calculation is a scalar value, so The dimension is 1*2F, where F is Therefore It is 2F*1.

[0126] This application adds the semantics of the opposite edge to the original formula, and the formula is rewritten as:

[0127]

[0128] Among them, att ij It is an improved attention mechanism to distinguish the original attention ij The attention mechanism makes the distinction; is a 2*2F weight matrix for extracting vertex eigenvalues. ij refers to the edge feature; LeakyReLU is the LeakyReLU function.

[0129] It means concatenating point features with edge features. The role of The function of is the same as that of , which is a 1*4 weight matrix used to calculate the scalar value, namely the attention score. Through this improvement, this application also includes the characteristics of the edge into the assessment scope of the attention mechanism. This is a very important improvement. In the blockchain network, V i If you give V j Transferring a large sum of money or transferring money frequently. This is an obvious behavioral feature, and attention should be paid to this situation. However, the traditional attention mechanism does not take this factor into account.

[0130] In order to further improve the application intelligence of the blockchain transaction privacy protector penetration model, in a blockchain transaction privacy protector penetration model training method provided in an embodiment of the present application, see Figure 6 The graph attention neural network model in the blockchain transaction privacy protector penetration model training method also specifically includes the following contents:

[0131] A binary classifier connected to the cosine loss layer, the binary classifier is used to determine whether the cosine loss value output by the cosine loss layer is within a preset transaction behavior difference threshold range, and if so, output abnormal transaction identification result data indicating that there is no abnormality in the single address conversion transaction data; if not, output abnormal transaction identification result data indicating that there is an abnormality in the single address conversion transaction data.

[0132] It is understandable that the transaction behavior difference threshold range can be set according to actual application requirements, for example, it can be set to 0 to 0.3.

[0133] Based on the above-mentioned blockchain transaction privacy protector penetration model training method embodiment, this application also provides a blockchain transaction privacy protector penetration method embodiment, see Figure 7 The blockchain transaction privacy protector penetration method specifically includes the following contents:

[0134] Step 300: Obtain each single address conversion transaction data corresponding to the entity data obtained after being processed by the transaction privacy protector in the blockchain network, and respectively extract each directed graph feature data corresponding to each of the single address conversion transaction data.

[0135] Step 400: Input each directed graph feature data corresponding to each single address conversion transaction data corresponding to the entity data into the blockchain transaction privacy protector penetration model, so that the blockchain transaction privacy protector penetration model outputs each abnormal transaction identification result data corresponding to each single address conversion transaction data corresponding to the entity data, wherein the blockchain transaction privacy protector penetration model is pre-trained based on the blockchain transaction privacy protector penetration model training method.

[0136] It can be understood that the specific implementation methods of the model online application process from step 300 to step 400 of the present application can refer to the implementation methods of each step in the aforementioned blockchain transaction privacy protector penetration model training method. In addition, the blockchain transaction privacy protector penetration model mentioned in step 400 can also be implemented using the blockchain transaction privacy protector penetration model training method mentioned in the aforementioned embodiment, which will not be repeated here.

[0137] From the above description, it can be seen that the blockchain transaction privacy protector penetration method provided in the embodiment of the present application can realize the penetration of the blockchain transaction privacy protector based on the model, thereby improving the traceability and security of transactions executed by the transaction privacy protector.

[0138] In order to further improve the intelligent degree of penetration of the blockchain transaction privacy protector and the effectiveness of abnormal transaction identification result data, in the blockchain transaction privacy protector penetration method provided in the embodiment of the present application, see Figure 8 , the blockchain transaction privacy protector penetration method further specifically includes the following contents after step 400:

[0139] Step 500: From the abnormal transaction identification result data corresponding to each of the single address conversion transaction data corresponding to the entity data, the number of abnormal transaction identification result data indicating that the single address conversion transaction data does not have abnormality is obtained as the abnormal transaction value.

[0140] Step 600: Determine whether the ratio between the abnormal transaction value corresponding to the entity data and the total number of the single address conversion transaction data corresponding to the entity data is less than an abnormal transaction ratio threshold; if the ratio is less than the abnormal transaction ratio threshold and the ratio is 0, generate entity recognition result data indicating that the entity data has no abnormal risk; if the ratio is less than the abnormal transaction ratio threshold and the ratio is not 0, generate entity recognition result data indicating that the abnormal risk level of the entity data is a first risk level; if the ratio is equal to or greater than the abnormal transaction ratio threshold, generate entity recognition result data indicating that the abnormal risk level of the entity data is a second risk level.

[0141] It is understandable that the abnormal transaction ratio threshold can be set according to actual application requirements, for example, it can be set to any ratio between 0% and 50%.

[0142] From the software level, the present application also provides a blockchain transaction privacy protector penetration model training device for executing all or part of the blockchain transaction privacy protector penetration model training method, see Fig. 9 The blockchain transaction privacy protector penetration model training device specifically includes the following contents:

[0143] The training data processing module 10 is used to extract the directed graph feature data corresponding to each single address conversion transaction data corresponding to each entity data obtained after being processed by the transaction privacy protector in the blockchain network; wherein the entity data includes: the address transaction set of the starting transaction address corresponding to the transaction account and the address transaction set corresponding to each other transaction address generated after the starting transaction address is converted at least once by the transaction privacy protector, and the address transaction set includes each transaction data corresponding to the transaction address uniquely corresponding to the address transaction set;

[0144] The model training module 20 is used to train a graph attention neural network model based on each directed graph feature data corresponding to each of the single address conversion transaction data, so as to train the graph attention neural network model into a blockchain transaction privacy protector penetration model that outputs abnormal transaction identification result data corresponding to the single address conversion transaction data according to each of the directed graph feature data corresponding to the single address conversion transaction data.

[0145] The embodiment of the blockchain transaction privacy protector penetration model training device provided in the present application can be specifically used to execute the processing flow of the embodiment of the blockchain transaction privacy protector penetration model training method in the above-mentioned embodiment. Its functions will not be repeated here, and reference can be made to the detailed description of the embodiment of the blockchain transaction privacy protector penetration model training method.

[0146] The part of the blockchain transaction privacy protector penetration model training device that performs blockchain transaction privacy protector penetration model training can be completed in the server or client device. It can be selected according to the processing capacity of the client device and the limitations of the user's usage scenario. This application is not limited to this. If all operations are completed in the client device, the client device may also include a processor for specific processing of blockchain transaction privacy protector penetration model training.

[0147] The client device may have a communication module (i.e., a communication unit) that can communicate with a remote server to achieve data transmission with the server. The server may include a server on the task scheduling center side, and other implementation scenarios may also include a server on an intermediate platform, such as a server on a third-party server platform that has a communication link with the task scheduling center server. The server may include a single computer device, or a server cluster consisting of multiple servers, or a server structure of a distributed device.

[0148] The server and the client device may communicate with each other using any suitable network protocol, including network protocols that have not yet been developed on the date of filing this application. The network protocols may include, for example, TCP / IP, UDP / IP, HTTP, HTTPS, etc. Of course, the network protocols may also include, for example, RPC (Remote Procedure Call Protocol) and REST (Representational State Transfer) protocols used on top of the above protocols.

[0149] From the above description, it can be seen that the blockchain transaction privacy protector penetration model training device provided in the embodiment of the present application can train a model for identifying whether there are anomalies in the transaction data processed by the blockchain transaction privacy protector, and then can realize the penetration of the blockchain transaction privacy protector based on the model, thereby improving the traceability and security of transactions executed by the transaction privacy protector.

[0150] From the software level, the present application also provides a blockchain transaction privacy protector penetration device for executing all or part of the blockchain transaction privacy protector penetration method, see Fig.10 The blockchain transaction privacy protector penetration device specifically includes the following contents:

[0151] The application data processing module 30 is used to obtain each single address conversion transaction data corresponding to the entity data obtained after processing by the transaction privacy protector in the blockchain network, and respectively extract each directed graph feature data corresponding to each of the single address conversion transaction data.

[0152] The model identification module 40 is used to input each directed graph feature data corresponding to each single address conversion transaction data corresponding to the entity data into the blockchain transaction privacy protector penetration model, so that the blockchain transaction privacy protector penetration model outputs each abnormal transaction identification result data corresponding to each single address conversion transaction data corresponding to the entity data, wherein the blockchain transaction privacy protector penetration model is pre-trained based on the blockchain transaction privacy protector penetration model training method.

[0153] The embodiment of the blockchain transaction privacy protector penetration device provided in this application can be specifically used to execute the processing flow of the embodiment of the blockchain transaction privacy protector penetration method in the above-mentioned embodiment. Its functions will not be repeated here, and reference can be made to the detailed description of the above-mentioned blockchain transaction privacy protector penetration method embodiment.

[0154] The part of the blockchain transaction privacy protector penetration device that performs blockchain transaction privacy protector penetration can be completed in the server or client device. It can be selected according to the processing capacity of the client device and the limitations of the user's usage scenario. This application does not limit this. If all operations are completed in the client device, the client device may also include a processor for specific processing of blockchain transaction privacy protector penetration.

[0155] The embodiment of the present application also provides an electronic device, which may include a processor, a memory, a receiver and a transmitter, wherein the processor is used to execute the blockchain transaction privacy protector penetration model training method and / or the blockchain transaction privacy protector penetration method mentioned in the above embodiment. The processor and the memory may be connected via a bus or other means, taking the bus connection as an example. The receiver may be connected to the processor and the memory via a wired or wireless manner.

[0156] The processor may be a central processing unit (CPU). The processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or a combination of the above chips.

[0157] As a non-transient computer-readable storage medium, the memory can be used to store non-transient software programs, non-transient computer executable programs and modules, such as the blockchain transaction privacy protector penetration model training method and / or the blockchain transaction privacy protector penetration method in the embodiment of the present application. The processor executes various functional applications and data processing of the processor by running the non-transient software programs, instructions and modules stored in the memory, that is, the blockchain transaction privacy protector penetration model training method and / or the blockchain transaction privacy protector penetration method in the above method embodiment is implemented.

[0158] The memory may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created by the processor, etc. In addition, the memory may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the memory may optionally include a memory remotely arranged relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0159] The one or more modules are stored in the memory, and when executed by the processor, execute the blockchain transaction privacy protector penetration model training method and / or the blockchain transaction privacy protector penetration method in the embodiment.

[0160] In some embodiments of the present application, the user equipment may include a processor, a memory, and a transceiver unit, which may include a receiver and a transmitter. The processor, memory, receiver, and transmitter may be connected through a bus system. The memory is used to store computer instructions, and the processor is used to execute the computer instructions stored in the memory to control the transceiver unit to send and receive signals.

[0161] As an implementation method, the functions of the receiver and the transmitter in the present application can be considered to be implemented through a transceiver circuit or a dedicated chip for transceiver, and the processor can be considered to be implemented through a dedicated processing chip, a processing circuit or a general chip.

[0162] As another implementation method, it is possible to use a general-purpose computer to implement the server provided in the embodiment of the present application, that is, to store the program code for implementing the functions of the processor, receiver, and transmitter in a memory, and the general-purpose processor implements the functions of the processor, receiver, and transmitter by executing the code in the memory.

[0163] The embodiment of the present application also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the steps of the aforementioned blockchain transaction privacy protector penetration model training method and / or the blockchain transaction privacy protector penetration method are implemented. The computer-readable storage medium can be a tangible storage medium, such as a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a floppy disk, a hard disk, a removable storage disk, a CD-ROM, or any other form of storage medium known in the technical field.

[0164] An embodiment of the present application also provides a computer program product, including a computer program, which, when executed by a processor, implements the blockchain transaction privacy protector penetration model training method and / or implements the blockchain transaction privacy protector penetration method.

[0165] It should be understood by those skilled in the art that the exemplary components, systems and methods described in conjunction with the embodiments disclosed herein can be implemented in hardware, software or a combination of the two. Whether it is specifically performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application. When implemented in hardware, it can be, for example, an electronic circuit, an application specific integrated circuit (ASIC), appropriate firmware, a plug-in, a function card, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. The program or code segment can be stored in a machine-readable medium, or transmitted on a transmission medium or a communication link via a data signal carried in a carrier.

[0166] It should be clear that the present application is not limited to the specific configuration and processing described above and shown in the figures. For the sake of simplicity, a detailed description of the known method is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present application is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of the present application.

[0167] In the present application, features described and / or illustrated for one embodiment may be used in the same manner or in a similar manner in one or more other embodiments, and / or combined with features of other embodiments or replace features of other embodiments.

[0168] The above description is only the preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the embodiments of the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A blockchain transaction privacy protector penetration model training method, characterized in that: include: For each single address conversion transaction data corresponding to each entity data obtained after being processed by the transaction privacy protector in the blockchain network, each directed graph feature data corresponding to each single address conversion transaction data is extracted respectively; wherein the entity data includes: an address transaction set of a starting transaction address corresponding to a transaction account and an address transaction set corresponding to each other transaction address generated after the starting transaction address is converted at least once by the transaction privacy protector, and the address transaction set includes each transaction data corresponding to the transaction address uniquely corresponding to the address transaction set; A graph attention neural network model is trained based on each directed graph feature data corresponding to each of the single address conversion transaction data, so that the graph attention neural network model is trained as a blockchain transaction privacy protector penetration model that outputs abnormal transaction identification result data corresponding to the single address conversion transaction data according to each of the directed graph feature data corresponding to the single address conversion transaction data.

2. The blockchain transaction privacy protector penetration model training method according to claim 1 is characterized in that: Before extracting the directed graph feature data corresponding to each of the single address conversion transaction data corresponding to each of the entity data obtained after being processed by the transaction privacy protector in the blockchain network, the method further includes: The real entity data corresponding to each transaction account obtained after being processed by the transaction privacy protector in the blockchain network are marked as positive samples respectively; and the new entity data generated after cross-construction of the address transaction set of the entity data corresponding to each transaction account is marked as a negative sample; Based on the number of times each of the entity data marked as the positive sample and the negative sample is subjected to address conversion by the transaction privacy protector, each of the entity data is split respectively to obtain each single address conversion transaction data corresponding to each of the entity data; wherein the single address conversion transaction data includes: a transaction set before address conversion corresponding to the source address and a transaction set after address conversion corresponding to the target address; the source address is any transaction address in the entity data, and the target address is another transaction address generated after the source address is subjected to an address conversion by the transaction privacy protector.

3. The blockchain transaction privacy protector penetration model training method according to claim 2 is characterized in that: The method extracts the directed graph feature data corresponding to each of the single address conversion transaction data obtained after being processed by the transaction privacy protector in the blockchain network, including: According to the transaction time sequence corresponding to each of the transaction data, based on a preset transaction quantity division threshold, the transaction set before address conversion and the transaction set after address conversion corresponding to each of the single address conversion transaction data are respectively divided into transaction data, so as to obtain directed graphs corresponding to the transaction set before address conversion and the transaction set after address conversion, wherein each of the directed graphs includes vertices for representing different transaction data and edges for representing transaction relationships between different transaction data; The graph feature data corresponding to each of the directed graphs are extracted respectively to obtain the directed graph feature data corresponding to each of the single address conversion transaction data; wherein the directed graph feature data includes a directed graph corresponding to the corresponding single address conversion transaction data and the graph feature data corresponding to the directed graph.

4. The blockchain transaction privacy protector penetration model training method according to claim 3 is characterized in that: The extracting graph feature data corresponding to each of the directed graphs respectively includes: The vertex feature data of each vertex and the edge feature data of each edge corresponding to each of the directed graphs are extracted respectively to form the graph feature data corresponding to each of the directed graphs.

5. The blockchain transaction privacy protector penetration model training method according to claim 4 is characterized in that: The graph attention neural network model includes: A graph attention neural network is used to extract features of the vertices and the edges in the directed graph for each directed graph feature data corresponding to the input single address conversion transaction data, so as to obtain graph readout embedding vectors corresponding to each directed graph feature data corresponding to the single address conversion transaction data; A first concatenation layer is used to concatenate graph-read embedding vectors corresponding to each directed graph feature data corresponding to the transaction set before address conversion, so as to obtain a concatenation vector corresponding to the transaction set before address conversion; The second concatenation layer is used to concatenate the graph read-out embedding vectors corresponding to the directed graph feature data corresponding to the post-address conversion transaction set, so as to obtain a concatenation vector corresponding to the post-address conversion transaction set; A first multi-layer perceptron is used to extract transaction behavior features from the concatenation vector corresponding to the pre-address conversion transaction set output by the first concatenation layer, so as to obtain pre-address conversion transaction behavior feature data corresponding to the pre-address conversion transaction set; A second multi-layer perceptron is used to extract transaction behavior features from the concatenated vector corresponding to the post-address conversion transaction set output by the second concatenated layer, so as to obtain post-address conversion transaction behavior feature data corresponding to the post-address conversion transaction set; And, a cosine loss layer is used to calculate the cosine loss value between the transaction behavior characteristic data before address conversion output by the first multi-layer perceptron and the transaction behavior characteristic data after address conversion output by the second multi-layer perceptron based on the cosine loss function.

6. The blockchain transaction privacy protector penetration model training method according to claim 5 is characterized in that: The graph attention neural network model also includes: A binary classifier is used to determine whether the cosine loss value output by the cosine loss layer is within a preset transaction behavior difference threshold range. If so, it outputs abnormal transaction identification result data indicating that there is no abnormality in the single address conversion transaction data; if not, it outputs abnormal transaction identification result data indicating that there is an abnormality in the single address conversion transaction data.

7. A blockchain transaction privacy protector penetration method, characterized in that: include: Obtaining each single address conversion transaction data corresponding to the entity data obtained after being processed by the transaction privacy protector in the blockchain network, and respectively extracting each directed graph feature data corresponding to each of the single address conversion transaction data; The directed graph feature data corresponding to each single address conversion transaction data corresponding to the entity data are input into the blockchain transaction privacy protector penetration model, so that the blockchain transaction privacy protector penetration model outputs the abnormal transaction identification result data corresponding to each single address conversion transaction data corresponding to the entity data, wherein the blockchain transaction privacy protector penetration model is pre-trained based on the blockchain transaction privacy protector penetration model training method described in any one of claims 1 to 6.

8. The blockchain transaction privacy protector penetration method according to claim 7, characterized in that: Also includes: From the abnormal transaction identification result data corresponding to each of the single address conversion transaction data corresponding to the entity data, the number of abnormal transaction identification result data indicating that the single address conversion transaction data does not have abnormality is obtained as the abnormal transaction value; Determine whether the ratio between the abnormal transaction value corresponding to the entity data and the total number of single address conversion transaction data corresponding to the entity data is less than an abnormal transaction ratio threshold; If the ratio is less than the abnormal transaction ratio threshold and the ratio is 0, generating entity recognition result data indicating that the entity data has no abnormal risk; If the ratio is less than the abnormal transaction ratio threshold and the ratio is not 0, generating entity recognition result data indicating that the abnormal risk level of the entity data is a first risk level; If the ratio is equal to or greater than the abnormal transaction ratio threshold, entity recognition result data indicating that the abnormal risk level of the entity data is a second risk level is generated.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, it implements the blockchain transaction privacy protector penetration model training method as described in any one of claims 1 to 6, and / or implements the blockchain transaction privacy protector penetration method as described in claim 7 or 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, it implements the blockchain transaction privacy protector penetration model training method as described in any one of claims 1 to 6, and / or implements the blockchain transaction privacy protector penetration method as described in claim 7 or 8.

Images