Block chain address risk comprehensive evaluation system based on weighted federal learning

Abstract

The invention discloses a block chain address risk comprehensive evaluation system based on weighted federal learning, and relates to the field of block chain address risk comprehensive evaluation, the block chain address risk comprehensive evaluation system comprises a central server, the central server comprises a global model, the central server is electrically connected with a coordinator, the coordinator is electrically connected with a participant, the participant comprises a local model, and the local model is electrically connected with the coordinator. The central server is electrically connected with a training unit. According to the block chain address risk comprehensive evaluation system based on weighted federated learning, an improved weighted federated learning method is used, the problem of data imbalance is solved on the basis of protecting user privacy and building a multi-party model, coordination convergence is carried out through models of different participants, a main model is updated, and the evaluation efficiency is improved. According to the risk comprehensive evaluation system provided by the invention, multi-party data can be effectively utilized, meanwhile, the privacy of the user is protected, and the user can be better helped to avoid risks existing in transactions.

Description

Technical Field

[0001] The present invention relates to the field of comprehensive risk assessment of blockchain addresses, and particularly to a comprehensive risk assessment system for blockchain addresses based on weighted federated learning. Background Art

[0002] A comprehensive risk assessment system for blockchain addresses is a technical system for evaluating risks related to blockchain addresses and wallet addresses. In the blockchain ecosystem, each participant can generate one or more wallet addresses for storing cryptocurrency assets or participating in transactions. However, these addresses may be subject to various risks and threats, including fraud, etc. Therefore, developing a comprehensive risk assessment system for blockchain addresses helps to identify and mitigate these risks. Common detection systems usually make risk judgments based on the following aspects: historical transaction behavior, rule metrics, artificial intelligence technology, real-time monitoring, and data integration.

[0003] Although machine learning technology can use massive data for training and effectively improve the accuracy of detection, the data quality usually determines the effectiveness of model detection. The data owned by a single institution is often limited. To use multi-party data sources, various problems such as privacy protection, data dispersion, data transmission, and multi-party collaboration are often faced. Therefore, federated learning technology can be used to solve these problems. Federated learning can be divided into horizontal federated learning and vertical federated learning according to the data dimension. The feature of horizontal federated learning is that different participants have data with the same features but different samples. The feature of vertical federated learning is that different participants have data with different features. However, the actual data situation may be more complex. Using general federated learning methods still has some defects: large multi-party communication overhead, low model synchronization efficiency, and data imbalance, etc.

[0004] Therefore, it is necessary to propose a comprehensive risk assessment system for blockchain addresses based on weighted federated learning to solve the above problems. Summary of the Invention

[0005] The main object of the present invention is to provide a comprehensive risk assessment system for blockchain addresses based on weighted federated learning, which can effectively solve the problems in the background art.

[0006] To achieve the above object, the technical solution adopted by the present invention is as follows:

[0007] A comprehensive risk assessment system for blockchain addresses based on weighted federated learning includes a central server, the central server includes a global model, the central server is electrically connected to a coordinator, the coordinator is electrically connected to participants, the participants include local models, and the central server is electrically connected to a training unit;

[0008] The central server is used for the layout of the basic and global models of the system. The training unit is used for privately training the downloaded models of other participating parties. The coordinator is used for system communication and data aggregation. The local model is used to detect the model status, aggregate the statuses of each model parameter, and update the main model.

[0009] Preferably, there are at least three participating parties. The output end of the coordinator is electrically connected to the input end of the central server, and the output end of the participating party is electrically connected to the input end of the coordinator.

[0010] Preferably, the participating party includes a user information input module. The input end of the user information input module is electrically connected to the output end of an encryption module. The output end of the user information input module is electrically connected to the input end of a model construction unit. The model construction unit includes a parameter status sorting module and an aggregation module. The output end of the model construction unit is electrically connected to the input end of an update module;

[0011] The user information input module is used for inputting information of the participating party users. The encryption module is used for encrypting the information input by the participating party users. The model construction unit constructs a model based on the input information. The parameter status sorting module is used for sorting the statuses of the model parameters. The aggregation module is used for aggregating the sorted model parameter statuses. The update module is used for updating the model constructed by the model construction unit to the main model, and the main model is the global model.

[0012] Preferably, the coordinator includes a system communication module and a data aggregation module. The input end of the system communication module is electrically connected to the output end of a line encryption module;

[0013] The system communication module is used for weighted averaging the model data information received from the participating parties and inputting it into the central server. The line encryption module is used for encrypting the input line of the system communication module. The data aggregation module is used for aggregating the data models of the participating parties and the local model.

[0014] Preferably, the training unit includes a contribution value threshold setting module and a weighted information import module. The output end of the weighted information import module is electrically connected to the input end of a data balancing processing module. The output end of the data balancing processing module is electrically connected to the input end of a model synchronization module. The output end of the model synchronization module is electrically connected to the input end of a data training AI module. The input end of the data training AI module is electrically connected to the output end of a reward module. The input end of the data training AI module is electrically connected to the output end of a transaction risk avoidance module;

[0015] The contribution value threshold setting module is used to set the threshold of the contribution value for the information data input by the participating parties. The contribution value threshold setting module is used to allocate weights to the participating parties and perform balancing processing. The weighted information import module is used to import weighted information for the data information input by the participating parties. The data balancing processing module is used to perform data balancing processing on the weighted information imported by different participating parties and data with large differences. The data balancing processing module considers the contribution of the global model. The data training AI module learns the data after balancing processing based on federated learning. The reward module is used to reward the training data results and perform success rate improvement processing.

[0016] Preferably, the logic layer of the central server includes a user interface, a risk assessment and analysis engine, a coordination layer, a data transmission and aggregation layer, and a data storage and management layer.

[0017] Beneficial effects

[0018] Compared with the prior art, the present invention provides a comprehensive evaluation system for blockchain address risks based on weighted federated learning, which has the following beneficial effects:

[0019] 1. The main purpose of the comprehensive evaluation system for blockchain address risks based on weighted federated learning is to coordinate multi-party data based on the method of weighted federated learning to more accurately analyze the risks of blockchain addresses. Weighted federated learning is an improved federated learning method that takes into account the contributions of different participating parties to the global model by allocating weights to them. Compared with ordinary federated learning, weighted federated learning has advantages, especially in cases where the data distribution is uneven or the importance of participating parties is different.

[0020] 2. The comprehensive evaluation system for blockchain address risks based on weighted federated learning uses an improved weighted federated learning method to solve the problem of data imbalance on the basis of protecting user privacy and multi-party model construction. In a blockchain transaction system, the data owned by different participating parties varies greatly. For the evaluation system, it is necessary to dynamically adjust the weights according to the actual situation, coordinate and converge through the models of different participating parties, update the main model, and perform the final calculation through the main model.

[0021] 3. The comprehensive evaluation system for blockchain address risks based on weighted federated learning. An ordinary blockchain risk detection system can only build a model using a single data source, while the comprehensive evaluation system for blockchain address risks based on weighted federated learning proposed by the present invention can effectively utilize multi-party data, enable more ecological participating parties to participate in model training and iteration, and at the same time protect user privacy. Using this platform can better help users avoid risks in transactions. Description of the drawings

[0022] Figure 1 It is the system diagram of the central server of the present invention;

[0023] Figure 2 It is the system diagram of the model building unit of the present invention;

[0024] Figure 3 It is the system diagram of the coordinator of the present invention;

[0025] Figure 4 It is the system diagram of the training unit of the present invention;

[0026] Figure 5 It is the logic diagram of the central server of the present invention. Detailed implementation manners

[0027] To make the technical means, creative features, achieved purposes and functions of the present invention easy to understand, the present invention will be further described below in conjunction with specific implementation manners.

[0028] Such as Figures 1-5As shown in the figure, a comprehensive evaluation system for blockchain address risks based on weighted federated learning includes a central server. The central server includes a global model. The central server is electrically connected to a coordinator, and the coordinator is electrically connected to participating parties. The participating parties include local models. The central server is electrically connected to a training unit. The central server is used for the layout of the system's foundation and global model. The training unit is used for privately training the issued models of other participating parties. The coordinator is used for system communication and data aggregation. The local model is used to detect the model status, aggregate the parameter status of each model, and update the main model. There are at least three participating parties. The output end of the coordinator is electrically connected to the input end of the central server, and the output end of the participating party is electrically connected to the input end of the coordinator. The participating party includes a user information input module. The input end of the user information input module is electrically connected to the output end of an encryption module. The output end of the user information input module is electrically connected to the input end of a model construction unit. The model construction unit includes a parameter status sorting module and an aggregation module. The output end of the model construction unit is electrically connected to the input end of an update module. The update module, the user information input module is used for inputting the information of the participating party users. The encryption module is used for encrypting the information input by the participating party users. The model construction unit constructs a model based on the input information. The parameter status sorting module is used for sorting the status of the model parameters. The aggregation module is used for aggregating the sorted model parameter status. The update module is used for updating the model constructed by the model construction unit to the main model. The main model is the global model. The coordinator includes a system communication module and a data aggregation module. The input end of the system communication module is electrically connected to the output end of a line encryption module. The system communication module is used for weighted averaging the model data information received from the participating parties and inputting it into the central server. The line encryption module is used for encrypting the input line of the system communication module. The data aggregation module is used for aggregating the data models of the participating parties and local models. The training unit includes a contribution value threshold setting module and a weighted information import module. The output end of the weighted information import module is electrically connected to the input end of a data balancing processing module. The output end of the data balancing processing module is electrically connected to the input end of a model synchronization module. The output end of the model synchronization module is electrically connected to the input end of a data training AI module. The input end of the data training AI module is electrically connected to the output end of a reward module. The input end of the data training AI module is electrically connected to the output end of a transaction risk avoidance module. The contribution value threshold setting module is used for setting the threshold of the contribution value of the information data input by the participating parties. The contribution value threshold setting module is used for allocating the weights of the participating parties and performing balancing processing. The weighted information import module is used for importing weighted information of the data information input by the participating parties. The data balancing processing module is used for performing data balancing processing on the weighted information imported by different participating parties and data with large differences. The data balancing processing module considers the contribution of the global model. The data training AI module learns the balanced data based on federated learning.The reward module is used to reward the training data results and improve the success rate. The logic layer of the central server includes a user interface, a risk assessment and analysis engine, a coordination layer, a data transmission aggregation layer, and a data storage and management layer.

[0029] The blockchain address risk comprehensive evaluation system based on weighted federated learning mainly aims to coordinate multi-party data based on the method of weighted federated learning to more accurately analyze the risks of blockchain addresses. Weighted federated learning is an improved federated learning method. By assigning weights to different participants and taking their contributions to the global model into account, compared with ordinary federated learning, weighted federated learning has advantages, especially in the case of uneven data distribution or different importance of participants. An improved weighted federated learning method is used to solve the problem of data imbalance on the basis of protecting user privacy and multi-party model construction. In the blockchain transaction system, the data owned by different participants varies greatly. For the evaluation system, it is necessary to dynamically adjust the weights according to the actual situation, coordinate and converge through the models of different participants, update the main model, and perform the final calculation through the main model. The ordinary blockchain risk detection system can only build a model using a single data source, while the blockchain address risk comprehensive evaluation system based on weighted federated learning proposed by the present invention can effectively utilize multi-party data, enable more ecological participants to join the model training and iteration, and at the same time protect user privacy. Using this platform can better help users avoid the risks existing in transactions.

[0030] It should be noted that the present invention is a blockchain address risk comprehensive evaluation system based on weighted federated learning. When used, the central server is used as the system foundation to deploy the global model. For other participants, the model is sent down for private data training. System communication and data aggregation are carried out through the coordinator to detect the model status, aggregate the status of each model parameter, and update the main model to obtain the result in the way of weighted average.

[0031] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art of this industry should understand that the present invention is not limited by the above embodiments. What is described in the above embodiments and the specification only illustrates the principle of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. A blockchain address risk comprehensive assessment system based on weighted federated learning, including a central server, characterized in that: The central server includes a global model, the central server is electrically connected to a coordinator, the coordinator is electrically connected to a participant, the participant includes a local model, and the central server is electrically connected to a training unit; The central server is used for the layout of the system foundation and global model, the training unit is used to perform private data training on the models issued by other participants, the coordinator is used for system communication and data aggregation, and the local model is used to detect the model status, aggregate the parameter status of each model, and update the main model.

2. The blockchain address risk comprehensive assessment system based on weighted federated learning according to claim 1 is characterized in that: There are at least three participants, the output end of the coordinator is electrically connected to the input end of the central server, and the output ends of the participants are electrically connected to the input end of the coordinator.

3. The blockchain address risk comprehensive assessment system based on weighted federated learning according to claim 1 is characterized in that: The participant includes a user information input module, the input end of the user information input module is electrically connected to the output end of the encryption module, the output end of the user information input module is electrically connected to the input end of the model construction unit, the model construction unit includes a parameter state sorting module and a convergence module, the output end of the model construction unit is electrically connected to the input end of the update module, the update module; The user information input module is used for inputting information of participating users, the encryption module is used for encrypting the information input by participating users, the model building unit builds the model based on the input information, the parameter state sorting module is used for sorting the state of model parameters, the aggregation module is used for aggregating the sorted model parameter states, and the update module is used for updating the model constructed by the model building unit to the main model, and the main model is a global model.

4. The blockchain address risk comprehensive assessment system based on weighted federated learning according to claim 1 is characterized in that: The coordinator includes a system communication module and a data aggregation module, and the input end of the system communication module is electrically connected to the output end of the line encryption module; The system communication module is used to perform weighted averaging of the model data information received from the participants and input it to the central server, the line encryption module is used to encrypt the input line of the system communication module, and the data aggregation module is used to aggregate the data models of the participants and the local model.

5. The blockchain address risk comprehensive assessment system based on weighted federated learning according to claim 1 is characterized in that: The training unit includes a contribution value threshold setting module and a weighted information importing module, the output end of the weighted information importing module is electrically connected to the input end of the data balancing processing module, the output end of the data balancing processing module is electrically connected to the input end of the model synchronization module, the output end of the model synchronization module is electrically connected to the input end of the data training AI module, the input end of the data training AI module is electrically connected to the output end of the reward module, and the input end of the data training AI module is electrically connected to the output end of the transaction risk avoidance module; The contribution value threshold setting module is used to set the contribution value threshold of the information data input by the participants, the contribution value threshold setting module is used to allocate the weights of the participants and perform balancing processing, the weighted information importing module is used to import weighted information of the data information input by the participants, the data balancing processing module is used to perform data balancing processing on the weighted information imported by different participants and on the data with large differences, the data balancing processing module considers the contribution of the global model, the data training AI module learns the balanced data based on federated learning, and the reward module is used to reward the training data results and improve the success rate.

6. The blockchain address risk comprehensive assessment system based on weighted federated learning according to claim 1 is characterized in that: The logic layer of the central server includes a user interface, a risk assessment and analysis engine, a coordination layer, a data transmission aggregation layer, and a data storage and management layer.

Images