User portrait-based fraud-related number identification method and system

By using a user profiling-based identification method, leveraging big data on communication number history and expert models of sensitive information related to fraud, preliminary user profiling, fraud analysis and prediction, and information verification are conducted. This solves the difficulty of identifying fraudulent numbers in telecommunications fraud and achieves accurate identification and timely early warning.

CN118945262APending Publication Date: 2024-11-12BEIJING WEIZHIXINYE TECH CO LTD
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202410915620.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-09
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

In the current technology, telecommunications fraud occurs frequently and is difficult to identify. It is difficult to obtain preliminary user profile data of communication numbers, conduct fraud analysis and prediction based on profile data, and verify sensitive information, resulting in low accuracy and poor timeliness of identification.

Method used

The user profiling-based identification method uses big data on communication number history to create a preliminary user profile. It then uses a profile data analysis and prediction model and an expert model for sensitive information related to fraud to perform fraud analysis, prediction, and information verification. Combined with dual-end early warning and linkage tracing, it can accurately identify fraudulent numbers.

Benefits of technology

It significantly improves the accuracy of identifying fraudulent phone numbers and the timeliness of early warnings, reduces verification errors, improves the efficiency of fraud tracing, and enables early detection of hidden or deceptive fraud risks.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118945262A_ABST
    Figure CN118945262A_ABST
Patent Text Reader

Abstract

The invention provides a fraud-related number identification method and system based on a user portrait, and the method comprises the steps: carrying out the preliminary user portrait of a communication number based on the historical record big data of the communication number, and obtaining the preliminary user portrait data of the communication number; according to the preliminary user portrait data of the communication number, performing portrait data fraud-related analysis and prediction through a portrait data analysis and prediction model, and obtaining portrait data fraud-related analysis and prediction information; performing information verification through the fraud-related sensitive information expert model according to the portrait data fraud-related analysis prediction information, and obtaining an information verification result of the fraud-related sensitive information expert model; and according to a fraud-related sensitive information expert model information verification result, through fraud-related direction intelligent determination, communication number accurate user portraying is carried out, the communication number portraying result is sent to a communication number supervision terminal and a communication number receiving terminal, and double-terminal early warning linkage tracing is carried out to accurately identify the fraud-related number.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of information security intelligent identification technology, and more specifically, to a method and system for identifying fraudulent numbers based on user portraits. Background Art

[0002] At present, telecommunication fraud has become an increasingly serious threat to the social environment, and its frequent occurrence has made it increasingly difficult to identify fraudulent numbers. Specific issues include: how to obtain preliminary user portrait data of communication numbers, how to conduct fraud analysis and prediction of portrait data, how to verify expert model information of sensitive information related to fraud, how to conduct accurate user portraits of communication numbers, and how to accurately identify fraudulent numbers through early warning linkage tracing. These issues have yet to be resolved. Therefore, it is necessary to propose a method and system for identifying fraudulent numbers based on user portraits to at least partially solve the problems existing in the prior art. Summary of the invention

[0003] A series of simplified concepts are introduced in the content of the invention, which will be further described in detail in the specific implementation method section; the content of the invention of the present invention does not mean to attempt to limit the key features and essential technical features of the technical solution claimed for protection, nor does it mean to attempt to determine the scope of protection of the technical solution claimed for protection.

[0004] In order to at least partially solve the above problems, the present invention provides a method for identifying fraudulent numbers based on user portraits, comprising:

[0005] S100, based on the communication number historical record big data, perform preliminary user profiling on the communication number, and obtain preliminary user profiling data of the communication number;

[0006] S200, based on the preliminary user portrait data of the communication number, performing a portrait data fraud analysis and prediction through a portrait data analysis and prediction model to obtain the portrait data fraud analysis and prediction information;

[0007] S300, based on the fraud analysis prediction information of the portrait data, verify the information through the expert model of fraud-related sensitive information, and obtain the information verification result of the expert model of fraud-related sensitive information;

[0008] S400, based on the verification results of the expert model of sensitive information related to fraud, makes an intelligent judgment on the direction of fraud, and makes an accurate user portrait of the communication number. It sends the portrait to the communication number supervision end and the communication number receiving end, and conducts dual-end early warning linkage tracing to accurately identify the fraudulent number.

[0009] Preferably, S100 includes:

[0010] S101, extracting communication number history record user features based on communication number history record big data;

[0011] S102, creating a communication number history restoration service scenario by restoring the communication number service scenario according to the communication number history record user characteristics;

[0012] S103, restore the business scenario according to the communication number history, create the communication number historical behavior pattern through behavior pattern analysis, make a preliminary user portrait of the communication number, and obtain the preliminary user portrait data of the communication number; the communication number historical behavior pattern analysis includes: communication number historical call record analysis, communication number historical SMS record analysis and communication number historical network interaction data analysis.

[0013] Preferably, S200 includes:

[0014] S201, creating a profile data analysis and prediction model based on the preliminary user profile data of the communication number; performing fraud analysis and prediction on the profile data through the profile data analysis and prediction model, and analyzing and predicting whether the preliminary user profile data has a risk of fraud;

[0015] S202, analyzing and predicting preliminary user portrait data output results according to the portrait data analysis and prediction model, and obtaining the portrait data fraud analysis and prediction information;

[0016] The portrait data analysis and prediction model includes: a portrait data reading unit, a portrait data processing, analysis and prediction unit and a prediction information output unit; the portrait data processing, analysis and prediction unit includes: a data rule configuration processing subunit, a portrait data labeling management subunit and a labeling data analysis and prediction subunit; the preliminary user portrait data of the communication number is read by the portrait data reading unit and input into the portrait data processing, analysis and prediction unit, and the data rule configuration processing subunit of the portrait data processing, analysis and prediction unit performs portrait data conversion and processing monitoring to obtain configuration conversion processing monitoring data; the portrait data labeling management subunit processes the monitoring data according to the configuration conversion through data labeling and data directory association management to obtain monitoring labeling associated data; the labeling data analysis and prediction subunit analyzes the monitoring labeling associated data to analyze whether the label association similarity between the monitoring labeling associated data and the labeling data of the fraudulent number reaches or exceeds the reference label association similarity of the fraudulent number; when the label association similarity reaches or exceeds the reference label association similarity of the fraudulent number, it is predicted that the preliminary user portrait data has a risk of fraud; the prediction information output unit outputs the preliminary user portrait data with a risk of fraud.

[0017] Preferably, S300 includes:

[0018] S301, based on the collection of fraud-related sensitive information, an expert model of fraud-related sensitive information is created through intensive training of a fraud-related sensitive information integrated decision tree;

[0019] S302, verifying the fraud-related sensitive information of the image data fraud-related analysis and prediction information through the fraud-related sensitive information expert model;

[0020] S303, based on the fraud-related sensitive information verification, screen out the results having fraud-related sensitive information after verification, and obtain the fraud-related sensitive information expert model information verification result;

[0021] The expert model of sensitive information related to fraud is created by strengthening the integrated decision tree of sensitive information related to fraud, including: extracting the metric features of sensitive information related to fraud, and marking the sensitive attributes of all metric features related to fraud; extracting global information in sensitive information related to fraud, dividing global data into multiple subsets, obtaining a set of rules related to sensitive information related to fraud, and using them as nodes of the integrated decision tree to form a first decision path, and dividing the training samples of sensitive information related to fraud into first training samples of sensitive information related to fraud and second training samples of sensitive information related to fraud; expanding the first decision path according to the sensitive attributes related to fraud, and obtaining a second decision path; adding the first decision rule and the second decision rule to the first child node and the second child node of the integrated decision tree node, respectively, and the first decision rule and the second decision rule are reused multiple times; according to the first decision rule, the first child node branch is divided into the first training sample score of sensitive information related to fraud and the first child node decision rule; according to the second decision rule, the second child node branch is divided into the second training sample score of sensitive information related to fraud and the second child node decision rule; the first child node decision rule and the second child node decision rule are branched in sequence until the tree leaf node of the integrated decision tree is reached, and all the tree leaf nodes of the integrated decision tree are the scores of the training samples of sensitive information related to fraud; calculating the final sample score by weighted average; and creating the expert model of sensitive information related to fraud.

[0022] Preferably, S400 includes:

[0023] S401, based on the information verification result of the expert model of the fraud-related sensitive information, intelligently determine the fraud-related direction of the fraud-related sensitive information;

[0024] S402, through intelligent determination of the fraudulent direction, accurate user profiling of the fraudulent number is performed, and the fraudulent number is accurately identified;

[0025] S403, sending the accurate user portrait of the fraudulent number to the communication number supervision terminal and the communication number receiving terminal, and performing double-end early warning linkage tracing and accurate identification of the fraudulent number;

[0026] The precise user portrait of the fraudulent number is sent to the communication number supervision end and the communication number receiving end, and the fraudulent number is accurately identified through dual-end early warning linkage, including: sending the precise user portrait of the fraudulent number to the communication number supervision end and the communication number receiving end; at the communication number receiving end, first sending fraudulent early warning information, and alerting the communication number receiving end that its communication method has been obtained by the fraudulent number end, and conducting dual-end linkage early warning with the communication number supervision end, tracing the communication method leakage path, accurately identifying the fraudulent number to obtain sensitive fraud information channels, and conducting dual-end early warning linkage tracing to accurately identify the fraudulent number.

[0027] The present invention provides a fraudulent number identification system based on user portrait, including:

[0028] The big data preliminary portrait subsystem performs preliminary user portraits of communication numbers based on the big data of communication number history records, and obtains preliminary user portrait data of communication numbers;

[0029] The portrait data fraud analysis and prediction subsystem performs portrait data fraud analysis and prediction based on the preliminary user portrait data of the communication number through the portrait data analysis and prediction model to obtain the portrait data fraud analysis and prediction information;

[0030] The fraud-related information expert model verification subsystem predicts information based on the fraud-related analysis of the portrait data, verifies the information through the expert model of fraud-related sensitive information, and obtains the information verification results of the expert model of fraud-related sensitive information;

[0031] The fraud-related direction tracing and identification subsystem, based on the information verification results of the expert model of fraud-related sensitive information, makes an intelligent judgment on the fraud-related direction, and makes an accurate user profile of the communication number. The profile is sent to the communication number supervision end and the communication number receiving end, and a two-end early warning linkage tracing is performed to accurately identify the fraud-related number.

[0032] Preferably, the big data preliminary portrait subsystem includes:

[0033] User feature extraction subsystem, extracting user features of communication number history records based on communication number history record big data;

[0034] The number business scenario restoration subsystem creates a communication number history restoration business scenario through communication number business scenario restoration based on the communication number history record user characteristics;

[0035] The behavioral analysis preliminary profiling subsystem restores the business scenario based on the communication number history, creates the communication number historical behavior pattern through behavioral pattern analysis, makes a preliminary user profiling of the communication number, and obtains the preliminary user profiling data of the communication number; the communication number historical behavior pattern analysis includes: communication number historical call record analysis, communication number historical SMS record analysis and communication number historical network interaction data analysis.

[0036] Preferably, the image data fraud analysis and prediction subsystem includes:

[0037] The prediction model fraud analysis subsystem creates a profile data analysis prediction model based on the preliminary user profile data of the communication number; performs profile data fraud analysis and prediction through the profile data analysis prediction model, and analyzes and predicts whether the preliminary user profile data has fraud risks;

[0038] The fraud prediction information output subsystem analyzes and predicts the preliminary user portrait data output results based on the portrait data analysis and prediction model, and obtains the portrait data fraud analysis and prediction information;

[0039] The portrait data analysis and prediction model includes: a portrait data reading unit, a portrait data processing, analysis and prediction unit and a prediction information output unit; the portrait data processing, analysis and prediction unit includes: a data rule configuration processing subunit, a portrait data labeling management subunit and a labeling data analysis and prediction subunit; the preliminary user portrait data of the communication number is read by the portrait data reading unit and input into the portrait data processing, analysis and prediction unit, and the data rule configuration processing subunit of the portrait data processing, analysis and prediction unit performs portrait data conversion and processing monitoring to obtain configuration conversion processing monitoring data; the portrait data labeling management subunit processes the monitoring data according to the configuration conversion through data labeling and data directory association management to obtain monitoring labeling associated data; the labeling data analysis and prediction subunit analyzes the monitoring labeling associated data to analyze whether the label association similarity between the monitoring labeling associated data and the labeling data of the fraudulent number reaches or exceeds the reference label association similarity of the fraudulent number; when the label association similarity reaches or exceeds the reference label association similarity of the fraudulent number, it is predicted that the preliminary user portrait data has a risk of fraud; the prediction information output unit outputs the preliminary user portrait data with a risk of fraud.

[0040] Preferably, the fraud-related information expert model verification subsystem includes:

[0041] The fraud-related sensitive information expert model subsystem creates a fraud-related sensitive information expert model through fraud-related sensitive information integrated decision tree intensive training based on the collection of fraud-related sensitive information;

[0042] The fraud-related sensitive information verification subsystem verifies the fraud-related sensitive information of the image data fraud analysis and prediction information through the fraud-related sensitive information expert model;

[0043] The information verification result screening subsystem verifies the sensitive information related to fraud, screens out the results with the sensitive information related to fraud after verification, and obtains the information verification results of the expert model of the sensitive information related to fraud;

[0044] The expert model of sensitive information related to fraud is created by strengthening the integrated decision tree of sensitive information related to fraud, including: extracting the metric features of sensitive information related to fraud, and marking the sensitive attributes of all metric features related to fraud; extracting global information in sensitive information related to fraud, dividing global data into multiple subsets, obtaining a set of rules related to sensitive information related to fraud, and using them as nodes of the integrated decision tree to form a first decision path, and dividing the training samples of sensitive information related to fraud into first training samples of sensitive information related to fraud and second training samples of sensitive information related to fraud; expanding the first decision path according to the sensitive attributes related to fraud, and obtaining a second decision path; adding the first decision rule and the second decision rule to the first child node and the second child node of the integrated decision tree node, respectively, and the first decision rule and the second decision rule are reused multiple times; according to the first decision rule, the first child node branch is divided into the first training sample score of sensitive information related to fraud and the first child node decision rule; according to the second decision rule, the second child node branch is divided into the second training sample score of sensitive information related to fraud and the second child node decision rule; the first child node decision rule and the second child node decision rule are branched in sequence until the tree leaf node of the integrated decision tree is reached, and all the tree leaf nodes of the integrated decision tree are the scores of the training samples of sensitive information related to fraud; calculating the final sample score by weighted average; and creating the expert model of sensitive information related to fraud.

[0045] Preferably, the fraud-related direction tracing identification subsystem includes:

[0046] The fraud-related direction intelligent judgment subsystem makes an intelligent judgment on the fraud-related direction of the sensitive information based on the information verification results of the expert model of the sensitive information involved in fraud;

[0047] The subsystem for accurate user profiling of fraudulent numbers can accurately identify fraudulent numbers through intelligent judgment of the fraudulent direction and accurate user profiling of fraudulent numbers;

[0048] The dual-end early warning tracing and identification subsystem sends the accurate user portrait of the fraudulent number to the communication number supervision end and the communication number receiving end, and conducts dual-end early warning linkage tracing and accurate identification of the fraudulent number;

[0049] The precise user portrait of the fraudulent number is sent to the communication number supervision end and the communication number receiving end, and the fraudulent number is accurately identified through dual-end early warning linkage, including: sending the precise user portrait of the fraudulent number to the communication number supervision end and the communication number receiving end; at the communication number receiving end, first sending fraudulent early warning information, and alerting the communication number receiving end that its communication method has been obtained by the fraudulent number end, and conducting dual-end linkage early warning with the communication number supervision end, tracing the communication method leakage path, accurately identifying the fraudulent number to obtain sensitive fraud information channels, and conducting dual-end early warning linkage tracing to accurately identify the fraudulent number.

[0050] Compared with the prior art, the present invention has at least the following beneficial effects:

[0051] The present invention provides a method and system for identifying fraudulent numbers based on user portraits, including: based on the big data of communication number historical records, preliminary user portraits are made for communication numbers to obtain preliminary user portrait data for communication numbers; based on the preliminary user portrait data for communication numbers, fraud analysis and prediction of portrait data is performed through a portrait data analysis and prediction model to obtain fraud analysis and prediction information of portrait data; based on the fraud analysis and prediction information of portrait data, information verification is performed through an expert model of fraudulent sensitive information to obtain information verification results of the expert model of fraudulent sensitive information; based on the information verification results of the expert model of fraudulent sensitive information, accurate user portraits of communication numbers are made through intelligent judgment of fraudulent direction, and sent to the communication number supervision end and At the receiving end of communication numbers, two-way early warning linkage tracing is carried out to accurately identify fraudulent numbers; it can accurately identify fraudulent numbers based on the frequent and frequent characteristics of telecommunications fraud that seriously endangers the social environment; it can obtain preliminary user portrait data of communication numbers and obtain preliminary user portraits of fraudulent numbers on a large scale, and conduct fraud analysis and prediction of portrait data. In view of the difficulty in verifying user portraits, expert model information verification of fraud-related sensitive information can be carried out to significantly reduce verification errors; it can conduct accurate user portraits of communication numbers and early warning linkage tracing to accurately identify fraudulent numbers, greatly increase the accuracy of fraudulent number identification, significantly improve the timeliness of fraud-related early warnings, and significantly improve the efficiency of fraud tracing by timely tracing the root causes of fraudulent information channels.

[0052] The method and system for identifying fraudulent numbers based on user portraits described in the present invention, and other advantages, objectives and features of the present invention will be reflected in part through the following description, and in part will be understood by technical personnel in the field through research and practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0054] Figure 1 This is a diagram of an implementation scheme of the system for identifying fraudulent numbers based on user portraits described in the present invention.

[0055] Figure 2 This is a diagram of an embodiment of the fraudulent number identification system based on user portrait described in the present invention.

[0056] Figure 3 This is a diagram of an implementation step of the method for identifying fraudulent numbers based on user portraits described in the present invention. DETAILED DESCRIPTION

[0057] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments so that those skilled in the art can implement the invention according to the instructions. Figure 1-Figure 3As shown, the present invention provides a method for identifying fraudulent numbers based on user portraits, including:

[0058] S100, based on the communication number historical record big data, perform preliminary user profiling on the communication number, and obtain preliminary user profiling data of the communication number;

[0059] S200, based on the preliminary user portrait data of the communication number, performing a portrait data fraud analysis and prediction through a portrait data analysis and prediction model to obtain the portrait data fraud analysis and prediction information;

[0060] S300, based on the fraud analysis prediction information of the portrait data, verify the information through the expert model of fraud-related sensitive information, and obtain the information verification result of the expert model of fraud-related sensitive information;

[0061] S400, based on the verification results of the expert model of sensitive information related to fraud, makes an intelligent judgment on the direction of fraud, and makes an accurate user portrait of the communication number. It sends the portrait to the communication number supervision end and the communication number receiving end, and conducts dual-end early warning linkage tracing to accurately identify the fraudulent number.

[0062] The principle and effect of the above technical solution are as follows: The present invention provides a method for identifying fraudulent numbers based on user portraits, including: based on the big data of historical records of communication numbers, preliminary user portraits of communication numbers are performed to obtain preliminary user portrait data of communication numbers; based on the preliminary user portrait data of communication numbers, fraud analysis and prediction of portrait data are performed through a portrait data analysis and prediction model to obtain fraud analysis prediction information of portrait data; based on the fraud analysis prediction information of portrait data, information verification is performed through an expert model of fraud-related sensitive information to obtain information verification results of the expert model of fraud-related sensitive information; based on the information verification results of the expert model of fraud-related sensitive information, accurate user portraits of communication numbers are performed through intelligent judgment of the fraud-related direction, and sent to the communication number. The code supervision end and the communication number receiving end can conduct two-way early warning linkage tracing to accurately identify fraudulent numbers; it can accurately identify fraudulent numbers in view of the frequent and frequent characteristics of telecommunications fraud that seriously endangers the social environment; it can obtain preliminary user portrait data of communication numbers and obtain preliminary user portraits of fraudulent numbers on a large scale, and can conduct fraud analysis and prediction of portrait data. In view of the difficulty in user portrait verification, it can conduct information verification of fraud-related sensitive information expert model, which significantly reduces verification errors; it can conduct accurate user portraits of communication numbers and early warning linkage tracing to accurately identify fraudulent numbers, greatly increase the accuracy of fraudulent number identification, significantly improve the timeliness of fraud warnings, and significantly improve the efficiency of fraud tracing by timely tracing the root causes of fraudulent information channels.

[0063] In one embodiment, S100 includes:

[0064] S101, extracting communication number history record user features based on communication number history record big data;

[0065] S102, creating a communication number history restoration service scenario by restoring the communication number service scenario according to the communication number history record user characteristics;

[0066] S103, restore the business scenario according to the communication number history, create the communication number historical behavior pattern through behavior pattern analysis, make a preliminary user portrait of the communication number, and obtain the preliminary user portrait data of the communication number; the communication number historical behavior pattern analysis includes: communication number historical call record analysis, communication number historical SMS record analysis and communication number historical network interaction data analysis.

[0067] The principles and effects of the above technical solution are: based on the big data of communication number history records, user characteristics of communication number history records are extracted; according to the user characteristics of communication number history records, through the communication number business scenario restoration, a communication number history restoration business scenario is created; according to the communication number history restoration business scenario, through the behavior pattern analysis, a communication number history behavior pattern is created, a preliminary user portrait of the communication number is made, and preliminary user portrait data of the communication number is obtained; the communication number historical behavior pattern analysis includes: analysis of communication number historical call records, communication number historical SMS records and communication number historical network interaction data; preliminary user portrait data of communication numbers can be obtained to obtain preliminary user portraits of fraudulent numbers on a large scale; preliminary user portrait data of communication numbers can be obtained to obtain preliminary user portraits of fraudulent numbers on a large scale, greatly improving data acquisition efficiency and the breadth of data coverage.

[0068] In one embodiment, S200 includes:

[0069] S201, creating a profile data analysis and prediction model based on the preliminary user profile data of the communication number; performing fraud analysis and prediction on the profile data through the profile data analysis and prediction model, and analyzing and predicting whether the preliminary user profile data has a risk of fraud;

[0070] S202, analyzing and predicting preliminary user portrait data output results according to the portrait data analysis and prediction model, and obtaining the portrait data fraud analysis and prediction information;

[0071] The portrait data analysis and prediction model includes: a portrait data reading unit, a portrait data processing, analysis and prediction unit and a prediction information output unit; the portrait data processing, analysis and prediction unit includes: a data rule configuration processing subunit, a portrait data labeling management subunit and a labeling data analysis and prediction subunit; the preliminary user portrait data of the communication number is read by the portrait data reading unit and input into the portrait data processing, analysis and prediction unit, and the data rule configuration processing subunit of the portrait data processing, analysis and prediction unit performs portrait data conversion and processing monitoring to obtain configuration conversion processing monitoring data; the portrait data labeling management subunit processes the monitoring data according to the configuration conversion through data labeling and data directory association management to obtain monitoring labeling associated data; the labeling data analysis and prediction subunit analyzes the monitoring labeling associated data to analyze whether the label association similarity between the monitoring labeling associated data and the labeling data of the fraudulent number reaches or exceeds the reference label association similarity of the fraudulent number; when the label association similarity reaches or exceeds the reference label association similarity of the fraudulent number, it is predicted that the preliminary user portrait data has a risk of fraud; the prediction information output unit outputs the preliminary user portrait data with a risk of fraud.

[0072] The principle and effect of the above technical solution are as follows: based on the preliminary user portrait data of the communication number, a portrait data analysis and prediction model is created; the portrait data fraud analysis and prediction is performed through the portrait data analysis and prediction model to analyze and predict whether the preliminary user portrait data has a risk of fraud; the preliminary user portrait data output result is analyzed and predicted according to the portrait data analysis and prediction model to obtain the portrait data fraud analysis and prediction information; the portrait data analysis and prediction model includes: a portrait data reading unit, a portrait data processing analysis and prediction unit and a prediction information output unit; the portrait data processing analysis and prediction unit includes: a data rule configuration processing subunit, a portrait data labeling management subunit and a labeling data analysis and prediction subunit; the preliminary user portrait data of the communication number is read by the portrait data reading unit and input into the portrait data processing analysis and prediction unit, and after the portrait data processing analysis and prediction The data rule configuration processing subunit of the unit performs portrait data conversion and processing monitoring to obtain configuration conversion processing monitoring data; the portrait data labeling management subunit obtains monitoring labeling related data through data labeling and data directory association management according to the configuration conversion processing monitoring data; the labeling data analysis and prediction subunit analyzes the monitoring labeling related data to analyze whether the label association similarity between the monitoring labeling related data and the fraud-related number labeling data reaches or exceeds the fraud-related reference label association similarity; when the label association similarity reaches or exceeds the fraud-related reference label association similarity, it is predicted that the preliminary user portrait data has a fraud-related risk; the prediction information output unit outputs the preliminary user portrait data with a fraud-related risk; the portrait data fraud analysis and prediction can be performed, so that the fraud-related risks that are hidden deeper or more confusing can be warned and discovered earlier.

[0073] In one embodiment, S300 includes:

[0074] S301, based on the collection of fraud-related sensitive information, an expert model of fraud-related sensitive information is created through intensive training of a fraud-related sensitive information integrated decision tree;

[0075] S302, verifying the fraud-related sensitive information of the image data fraud-related analysis and prediction information through the fraud-related sensitive information expert model;

[0076] S303, based on the fraud-related sensitive information verification, screen out the results having fraud-related sensitive information after verification, and obtain the fraud-related sensitive information expert model information verification result;

[0077] The expert model of sensitive information related to fraud is created by strengthening the integrated decision tree of sensitive information related to fraud, including: extracting the metric features of sensitive information related to fraud, and marking the sensitive attributes of all metric features related to fraud; extracting the global information in the sensitive information related to fraud, dividing the global data into multiple subsets, obtaining the rule set of sensitive information related to fraud, and forming the first decision path as the node of the integrated decision tree, dividing the training sample of sensitive information related to fraud into the first training sample of sensitive information related to fraud and the second training sample of sensitive information related to fraud; expanding the first decision path according to the sensitive attributes related to fraud, and obtaining the second decision path; adding the first decision rule and the second decision rule to the first child node and the second child node of the integrated decision tree node respectively, and the first decision rule and the second decision rule are reused multiple times; according to the first decision rule, the first child node branch is divided into the first training sample score of sensitive information related to fraud and the first child node decision rule; according to the second decision rule, the second child node branch is divided into the second training sample score of sensitive information related to fraud and the second child node decision rule; the first child node decision rule and the second child node decision rule are branched in sequence until the tree leaf node of the integrated decision tree is reached, and all the tree leaf nodes of the integrated decision tree are the scores of the training samples of sensitive information related to fraud; calculating the final sample score by weighted average; creating the expert model of sensitive information related to fraud;

[0078] Based on the portrait data fraud analysis and prediction information data, the random forest is trained globally; based on the different clusters obtained by division, the expert model of fraud-related sensitive information is iteratively assigned and strengthened to extract local information in the portrait data fraud analysis and prediction information data; an assignment method based on accuracy as a threshold is used to assign an expert model of fraud-related sensitive information to each subset; the data in each subset is used to optimize the corresponding expert model of fraud-related sensitive information by adjusting the order of node features and recalculating the cutting score under the rearranged features; for any sample, the probability value of belonging to each expert model of fraud-related sensitive information is calculated respectively, and according to the size of all probability values, the sample is reallocated to the subset corresponding to the expert model of fraud-related sensitive information with the largest calculated probability value; the strengthening and reconstruction process of the expert model of fraud-related sensitive information is repeated until the performance of the expert model of fraud-related sensitive information is optimal.

[0079] The principle and effect of the above technical solution are as follows: based on the collection of sensitive information related to fraud, an expert model of sensitive information related to fraud is created through intensive training of an integrated decision tree of sensitive information related to fraud; through the expert model of sensitive information related to fraud, the fraud analysis and prediction information of the portrait data is verified for sensitive information related to fraud; based on the verification of sensitive information related to fraud, the results with sensitive information related to fraud after verification are screened out, and the verification results of the expert model of sensitive information related to fraud are obtained; an expert model of sensitive information related to fraud is created through intensive training of an integrated decision tree of sensitive information related to fraud, including: extracting the metric features of sensitive information related to fraud, marking the sensitive attributes of all metric features related to fraud; extracting the global information in the sensitive information related to fraud, dividing the global data into The method is divided into multiple subsets, and a set of rules for sensitive information related to fraud is obtained, which is used as an integrated decision tree node to form a first decision path, and the training samples of sensitive information related to fraud are divided into first training samples of sensitive information related to fraud and second training samples of sensitive information related to fraud; the first decision path is expanded according to the sensitive attribute related to fraud, and the second decision path is obtained; the first child node and the second child node of the integrated decision tree node are respectively added with the first decision rule and the second decision rule, and the first decision rule and the second decision rule are reused multiple times; according to the first decision rule, the first child node branch is divided into the first training sample score of sensitive information related to fraud and the first child node decision rule; according to the second decision rule, the second child node branch is divided into The second fraud-related sensitive information training sample score and the second child node decision rule; the first child node decision rule and the second child node decision rule are branched in sequence until the integrated decision tree leaf node is reached, and the integrated decision tree leaf nodes are all fraud-related sensitive information training sample scores; the final sample score is calculated by weighted average; an expert model of fraud-related sensitive information is created; based on the portrait data fraud analysis prediction information data, a random forest is trained globally; based on the different clusters obtained by division, the expert model of fraud-related sensitive information is iteratively assigned and strengthened to extract local information from the portrait data fraud analysis prediction information data; an assignment method based on accuracy as a threshold is used to assign fraud-related sensitive information to each subset. Expert model of fraud-related sensitive information; use the data in each subset to optimize the corresponding expert model of fraud-related sensitive information by adjusting the order of node features and recalculating the cutting scores under the rearranged features; for any sample, calculate the probability value of belonging to each expert model of fraud-related sensitive information, and according to the size of all probability values, reallocate the sample to the subset corresponding to the expert model of fraud-related sensitive information with the largest calculated probability value; repeat the reinforcement and reconstruction process of the expert model of fraud-related sensitive information until the performance of the expert model of fraud-related sensitive information is optimal; for the problem of difficult user portrait verification, information verification of the expert model of fraud-related sensitive information can be performed to significantly reduce the verification error.

[0080] In one embodiment, S400 includes:

[0081] S401, based on the information verification result of the expert model of the fraud-related sensitive information, intelligently determine the fraud-related direction of the fraud-related sensitive information;

[0082] S402, through intelligent determination of the fraudulent direction, accurate user profiling of the fraudulent number is performed, and the fraudulent number is accurately identified;

[0083] S403, sending the accurate user portrait of the fraudulent number to the communication number supervision terminal and the communication number receiving terminal, and performing double-end early warning linkage tracing and accurate identification of the fraudulent number;

[0084] The accurate user portrait of the fraudulent number is sent to the communication number supervision end and the communication number receiving end, and the fraudulent number is accurately identified by the dual-end early warning linkage, including: sending the accurate user portrait of the fraudulent number to the communication number supervision end and the communication number receiving end; at the communication number receiving end, firstly, the fraudulent early warning information is sent, and the communication number receiving end is warned that its communication method has been obtained by the fraudulent number end, and a dual-end linkage early warning is carried out with the communication number supervision end, the communication method leakage path is traced, and the fraudulent number is accurately identified to obtain the fraudulent sensitive information channel, and the fraudulent number is accurately identified by dual-end early warning linkage tracing;

[0085] Intelligent determination of the fraud direction of sensitive information involving fraud includes: classifying existing fraud types through information identification based on sensitive information involving fraud, and identifying and predicting new fraud types; determining the first decision path of the fraud direction in the first decision path based on the existing fraud type, and if the first decision path determines that it is an existing fraud type, directly performing a user portrait of the existing fraud type of the fraud number involved; if the first decision path determines that it is not an existing fraud type, performing a determination through the second decision path until a new fraud type is determined to be identified, predicting the development trend of the fraud direction of the new fraud type, and performing a user portrait of the new fraud type and a portrait of the development trend of the new fraud type; identifying and predicting the direction of the new fraud type based on the user portrait of the new fraud type and the portrait of the development trend of the new fraud type; intelligent determination of the fraud direction includes: determination of the financial fraud direction, determination of the asset fraud direction, and determination of the abduction fraud direction. Determination or gambling fraud direction determination; financial fraud direction determination includes: when the sensitive information includes fund fraud sensitive information, transfer fraud sensitive information or loan fraud sensitive information, it is determined to be financial fraud direction; asset fraud direction determination includes: when the sensitive information includes real estate fraud sensitive information, company fraud sensitive information or mortgage fraud sensitive information, it is determined to be asset fraud direction; abduction fraud direction determination includes: when the sensitive information includes address fraud sensitive information, transportation fraud sensitive information or travel fraud sensitive information, it is determined to be abduction fraud direction; gambling fraud direction determination includes: when the sensitive information includes gambling fraud sensitive information, winning fraud sensitive information or gambling fraud sensitive information, it is determined to be gambling fraud direction; intelligent fraud direction determination is performed based on the existing fraud type directions and predicted new fraud type directions.

[0086] The principle and effect of the above technical solution are as follows: based on the information verification result of the expert model of sensitive information related to fraud, intelligent judgment is made on the fraud direction of sensitive information related to fraud; through intelligent judgment on the fraud direction, accurate user portraits of fraudulent numbers are made to accurately identify fraudulent numbers; accurate user portraits of fraudulent numbers are sent to the communication number supervision end and the communication number receiving end, and dual-end early warning linkage is used to accurately identify fraudulent numbers; accurate user portraits of fraudulent numbers are sent to the communication number supervision end and the communication number receiving end, and dual-end early warning linkage is used to accurately identify fraudulent numbers, including: sending accurate user portraits of fraudulent numbers to the communication number supervision end and the communication number receiving end; at the communication number receiving end, fraudulent warning information is first sent, It also warns the receiving end of the communication number that its communication method has been obtained by the fraudulent number end, and conducts a two-way linkage warning with the communication number supervision end to trace the communication method leakage path, accurately identify the fraudulent number to obtain sensitive fraud information channels, and conduct two-way warning linkage tracing to accurately identify the fraudulent number; intelligently determine the fraudulent direction of sensitive fraud information, including: classifying existing fraud types through information identification based on sensitive fraud information, and identifying and predicting new fraud types; based on existing fraud types, determine the first decision path of the fraud direction in the first decision path. If the first decision path determines that it is an existing fraud type, directly conduct a user profile of the existing fraud type of the fraudulent number; if the first decision path determines that it is not an existing fraud If the fraud type is determined, the second decision path is used for judgment until a new type of fraud is identified, the fraud direction development trend of the new type of fraud is predicted, and a new type of fraud user portrait and a new type of fraud direction development trend prediction portrait are made; according to the new type of fraud user portrait and the new type of fraud direction development trend prediction portrait, the new type of fraud direction is identified and predicted; the intelligent judgment of the fraud direction includes: judgment of the financial fraud direction, judgment of the asset fraud direction, judgment of the abduction fraud direction or judgment of the gambling fraud direction; the judgment of the financial fraud direction includes: when the sensitive information of the fraud includes the sensitive information of the capital fraud, the transfer fraud or the loan fraud, it is judged as financial fraud; The direction of financial fraud; the direction of asset fraud includes: when the sensitive information includes real estate fraud, company fraud or mortgage fraud, it is determined to be asset fraud; the direction of abduction fraud includes: when the sensitive information includes address fraud, traffic fraud or travel fraud, it is determined to be abduction fraud; the direction of gambling fraud includes: when the sensitive information includes gambling fraud, lottery fraud or gambling fraud, it is determined to be gambling fraud; intelligent judgment of fraud direction is performed based on the existing fraud type direction and the predicted new fraud type direction;It can accurately identify fraudulent numbers through accurate user profiling of communication numbers and early warning linkage tracing, greatly increasing the accuracy of fraudulent number identification, significantly improving the timeliness of fraudulent early warnings, and significantly improving the efficiency of fraud tracing by timely tracing the source of fraudulent information channels.

[0087] The present invention provides a fraudulent number identification system based on user portrait, including:

[0088] The big data preliminary portrait subsystem performs preliminary user portraits of communication numbers based on the big data of communication number history records, and obtains preliminary user portrait data of communication numbers;

[0089] The portrait data fraud analysis and prediction subsystem performs portrait data fraud analysis and prediction based on the preliminary user portrait data of the communication number through the portrait data analysis and prediction model to obtain the portrait data fraud analysis and prediction information;

[0090] The fraud-related information expert model verification subsystem predicts information based on the fraud-related analysis of the portrait data, verifies the information through the expert model of fraud-related sensitive information, and obtains the information verification results of the expert model of fraud-related sensitive information;

[0091] The fraud-related direction tracing and identification subsystem, based on the information verification results of the expert model of fraud-related sensitive information, makes an intelligent judgment on the fraud-related direction, and makes an accurate user profile of the communication number. The profile is sent to the communication number supervision end and the communication number receiving end, and a two-end early warning linkage tracing is performed to accurately identify the fraud-related number.

[0092] The principle and effect of the above technical solution are as follows: the present invention provides a fraudulent number identification system based on user portrait, including: a big data preliminary portrait subsystem, which performs preliminary user portrait of the communication number based on the communication number historical record big data, and obtains preliminary user portrait data of the communication number; a portrait data fraud analysis and prediction subsystem, which performs portrait data fraud analysis and prediction through a portrait data analysis and prediction model according to the preliminary user portrait data of the communication number, and obtains portrait data fraud analysis and prediction information; a fraud information expert model verification subsystem, which verifies information through a fraud sensitive information expert model according to the portrait data fraud analysis and prediction information, and obtains the information verification result of the fraud sensitive information expert model; a fraud direction tracing identification subsystem, which verifies the information of the fraud sensitive information expert model according to the information verification result of the fraud direction through the fraud sensitive information expert model Intelligent judgment can be made to carry out accurate user portraits of communication numbers, and the portraits can be sent to the communication number supervision end and the communication number receiving end for two-way early warning linkage tracing to accurately identify fraudulent numbers. It can accurately identify fraudulent numbers in view of the frequent and frequent occurrence of telecommunications fraud that seriously endangers the social environment. It can obtain preliminary user portrait data of communication numbers and obtain preliminary user portraits of fraudulent numbers on a large scale, and conduct fraud analysis and prediction of portrait data. In view of the difficulty in verifying user portraits, it can verify the information of expert models of sensitive fraud information, significantly reducing the verification error. It can carry out accurate user portraits of communication numbers and early warning linkage tracing to accurately identify fraudulent numbers, greatly increasing the accuracy of fraudulent number identification, significantly improving the timeliness of fraud warnings, and significantly improving the efficiency of fraud tracing by timely tracing back to the root causes of fraudulent information channels.

[0093] In one embodiment, the big data preliminary profiling subsystem includes:

[0094] User feature extraction subsystem, extracting user features of communication number history records based on communication number history record big data;

[0095] The number business scenario restoration subsystem creates a communication number history restoration business scenario through communication number business scenario restoration based on the communication number history record user characteristics;

[0096] The behavioral analysis preliminary profiling subsystem restores the business scenario based on the communication number history, creates the communication number historical behavior pattern through behavioral pattern analysis, makes a preliminary user profiling of the communication number, and obtains the preliminary user profiling data of the communication number; the communication number historical behavior pattern analysis includes: communication number historical call record analysis, communication number historical SMS record analysis and communication number historical network interaction data analysis.

[0097] The principles and effects of the above technical solution are as follows: the big data preliminary portrait subsystem includes: a user feature extraction subsystem, which extracts user features of communication number history records based on the communication number history record big data; a number business scenario restoration subsystem, which creates a communication number history restoration business scenario through communication number business scenario restoration based on the communication number history user features; a behavior analysis preliminary portrait subsystem, which creates a communication number history behavior pattern through behavior pattern analysis based on the communication number history restoration business scenario, makes a preliminary user portrait of the communication number, and obtains preliminary user portrait data of the communication number; the communication number history behavior pattern analysis includes: analysis of communication number historical call records, communication number historical SMS records and communication number historical network interaction data; can obtain preliminary user portrait data of communication numbers and preliminary user portraits of fraudulent numbers on a large scale; can obtain preliminary user portrait data of communication numbers and preliminary user portraits of fraudulent numbers on a large scale, greatly improving data acquisition efficiency and the breadth of data coverage.

[0098] In one embodiment, the fraud analysis and prediction subsystem of the image data includes:

[0099] The prediction model fraud analysis subsystem creates a profile data analysis prediction model based on the preliminary user profile data of the communication number; performs profile data fraud analysis and prediction through the profile data analysis prediction model, and analyzes and predicts whether the preliminary user profile data has fraud risks;

[0100] The fraud prediction information output subsystem analyzes and predicts the preliminary user portrait data output results based on the portrait data analysis and prediction model, and obtains the portrait data fraud analysis and prediction information;

[0101] The portrait data analysis and prediction model includes: a portrait data reading unit, a portrait data processing, analysis and prediction unit and a prediction information output unit; the portrait data processing, analysis and prediction unit includes: a data rule configuration processing subunit, a portrait data labeling management subunit and a labeling data analysis and prediction subunit; the preliminary user portrait data of the communication number is read by the portrait data reading unit and input into the portrait data processing, analysis and prediction unit, and the data rule configuration processing subunit of the portrait data processing, analysis and prediction unit performs portrait data conversion and processing monitoring to obtain configuration conversion processing monitoring data; the portrait data labeling management subunit processes the monitoring data according to the configuration conversion through data labeling and data directory association management to obtain monitoring labeling associated data; the labeling data analysis and prediction subunit analyzes the monitoring labeling associated data to analyze whether the label association similarity between the monitoring labeling associated data and the labeling data of the fraudulent number reaches or exceeds the reference label association similarity of the fraudulent number; when the label association similarity reaches or exceeds the reference label association similarity of the fraudulent number, it is predicted that the preliminary user portrait data has a risk of fraud; the prediction information output unit outputs the preliminary user portrait data with a risk of fraud.

[0102] The principle and effect of the above technical solution are as follows: the portrait data fraud analysis and prediction subsystem includes: a prediction model fraud analysis subsystem, which creates a portrait data analysis and prediction model based on the preliminary user portrait data of the communication number; performs portrait data fraud analysis and prediction through the portrait data analysis and prediction model, and analyzes and predicts whether the preliminary user portrait data has fraud risks; a fraud prediction information output subsystem, which analyzes and predicts the output results of the preliminary user portrait data according to the portrait data analysis and prediction model, and obtains the portrait data fraud analysis and prediction information; the portrait data analysis and prediction model includes: a portrait data reading unit, a portrait data processing analysis and prediction unit and a prediction information output unit; the portrait data processing analysis and prediction unit includes: a data rule configuration processing subunit, a portrait data labeling management subunit and a labeling data analysis and prediction subunit; the preliminary user portrait data of the communication number is read through the portrait data reading unit and input into the portrait data processing unit The data rule configuration processing subunit of the portrait data processing, analysis and prediction unit performs portrait data conversion and processing monitoring through the data rule configuration processing subunit to obtain the configuration conversion processing monitoring data; the portrait data labeling management subunit obtains the monitoring labeling related data through data labeling and data directory association management according to the configuration conversion processing monitoring data; the labeling data analysis and prediction subunit analyzes the monitoring labeling related data to analyze whether the label association similarity between the monitoring labeling related data and the fraud-related number labeling data reaches or exceeds the fraud-related reference label association similarity; when the label association similarity reaches or exceeds the fraud-related reference label association similarity, it is predicted that the preliminary user portrait data has a fraud-related risk; the prediction information output unit outputs the preliminary user portrait data with a fraud-related risk; the portrait data fraud analysis and prediction can be performed, so that the fraud-related risks that are hidden deeper or more confusing can be warned and discovered earlier.

[0103] In one embodiment, the fraud information expert model verification subsystem includes:

[0104] The fraud-related sensitive information expert model subsystem creates a fraud-related sensitive information expert model through fraud-related sensitive information integrated decision tree intensive training based on the collection of fraud-related sensitive information;

[0105] The fraud-related sensitive information verification subsystem verifies the fraud-related sensitive information of the image data fraud analysis and prediction information through the fraud-related sensitive information expert model;

[0106] The information verification result screening subsystem verifies the sensitive information related to fraud, screens out the results with the sensitive information related to fraud after verification, and obtains the information verification results of the expert model of the sensitive information related to fraud;

[0107] The expert model of sensitive information related to fraud is created by strengthening the integrated decision tree of sensitive information related to fraud, including: extracting the metric features of sensitive information related to fraud, and marking the sensitive attributes of all metric features related to fraud; extracting the global information in the sensitive information related to fraud, dividing the global data into multiple subsets, obtaining the rule set of sensitive information related to fraud, and forming the first decision path as the node of the integrated decision tree, dividing the training sample of sensitive information related to fraud into the first training sample of sensitive information related to fraud and the second training sample of sensitive information related to fraud; expanding the first decision path according to the sensitive attributes related to fraud, and obtaining the second decision path; adding the first decision rule and the second decision rule to the first child node and the second child node of the integrated decision tree node respectively, and the first decision rule and the second decision rule are reused multiple times; according to the first decision rule, the first child node branch is divided into the first training sample score of sensitive information related to fraud and the first child node decision rule; according to the second decision rule, the second child node branch is divided into the second training sample score of sensitive information related to fraud and the second child node decision rule; the first child node decision rule and the second child node decision rule are branched in sequence until the tree leaf node of the integrated decision tree is reached, and all the tree leaf nodes of the integrated decision tree are the scores of the training samples of sensitive information related to fraud; calculating the final sample score by weighted average; creating the expert model of sensitive information related to fraud;

[0108] Based on the portrait data fraud analysis and prediction information data, the random forest is trained globally; based on the different clusters obtained by division, the expert model of fraud-related sensitive information is iteratively assigned and strengthened to extract local information in the portrait data fraud analysis and prediction information data; an assignment method based on accuracy as a threshold is used to assign an expert model of fraud-related sensitive information to each subset; the data in each subset is used to optimize the corresponding expert model of fraud-related sensitive information by adjusting the order of node features and recalculating the cutting score under the rearranged features; for any sample, the probability value of belonging to each expert model of fraud-related sensitive information is calculated respectively, and according to the size of all probability values, the sample is reallocated to the subset corresponding to the expert model of fraud-related sensitive information with the largest calculated probability value; the strengthening and reconstruction process of the expert model of fraud-related sensitive information is repeated until the performance of the expert model of fraud-related sensitive information is optimal.

[0109] The principle and effect of the above technical solution are as follows: the fraud-related information expert model verification subsystem includes: the fraud-related sensitive information expert model subsystem, which creates a fraud-related sensitive information expert model through enhanced training of the fraud-related sensitive information integrated decision tree according to the collection of fraud-related sensitive information; the fraud-related sensitive information verification subsystem, which verifies the fraud-related sensitive information of the image data fraud-related analysis and prediction information through the fraud-related sensitive information expert model; the information verification result screening subsystem, which verifies the fraud-related sensitive information and screens out the fraud-related sensitive information results after verification, and obtains the information verification results of the fraud-related sensitive information expert model; the fraud-related sensitive information expert model is created through enhanced training of the fraud-related sensitive information integrated decision tree, including: extracting the degree of fraud-related sensitive information Quantitative features, label the fraud-related sensitive attributes of all metric features; extract global information from fraud-related sensitive information, divide the global data into multiple subsets, obtain the fraud-related sensitive information rule set, and use it as an integrated decision tree node to form a first decision path, and divide the fraud-related sensitive information training samples into first fraud-related sensitive information training samples and second fraud-related sensitive information training samples; expand the first decision path according to the fraud-related sensitive attributes to obtain the second decision path; add the first decision rule and the second decision rule to the first child node and the second child node of the integrated decision tree node respectively, and the first decision rule and the second decision rule are reused multiple times; according to the first decision rule, the first child node branch is divided into the first fraud-related sensitive information training sample score and the second fraud-related sensitive information training sample score. A sub-node decision rule; according to the second decision rule, the second sub-node branch is divided into the second fraud-related sensitive information training sample score and the second sub-node decision rule; the first sub-node decision rule and the second sub-node decision rule are branched in sequence until the integrated decision tree leaf node is reached, and the integrated decision tree leaf nodes are all fraud-related sensitive information training sample scores; the final sample score is calculated by weighted average; an expert model of fraud-related sensitive information is created; based on the portrait data, fraud-related analysis and prediction information data, a random forest is trained globally; based on the different clusters obtained by division, the allocation and reinforcement of the expert model of fraud-related sensitive information is iteratively performed to extract local information in the portrait data, fraud-related analysis and prediction information data; the accuracy rate is used as the threshold The allocation method allocates an expert model of fraud-related sensitive information to each subset; uses the data in each subset to optimize its corresponding expert model of fraud-related sensitive information by adjusting the order of node features and recalculating the cutting score under the rearranged features; for any sample, calculates the probability value of belonging to each expert model of fraud-related sensitive information, and reallocates the sample to the subset corresponding to the expert model of fraud-related sensitive information with the largest calculated probability value according to the size of all probability values; repeats the reinforcement and reconstruction process of the expert model of fraud-related sensitive information until the performance of the expert model of fraud-related sensitive information is optimal; in view of the problem of difficult user portrait verification, information verification of the expert model of fraud-related sensitive information can be performed to significantly reduce the verification error.

[0110] In one embodiment, the fraud-related direction tracing identification subsystem includes:

[0111] The fraud-related direction intelligent judgment subsystem makes an intelligent judgment on the fraud-related direction of the sensitive information based on the information verification results of the expert model of the sensitive information involved in fraud;

[0112] The subsystem for accurate user profiling of fraudulent numbers can accurately identify fraudulent numbers through intelligent judgment of the fraudulent direction and accurate user profiling of fraudulent numbers;

[0113] The dual-end early warning tracing and identification subsystem sends the accurate user portrait of the fraudulent number to the communication number supervision end and the communication number receiving end, and conducts dual-end early warning linkage tracing and accurate identification of the fraudulent number;

[0114] The accurate user portrait of the fraudulent number is sent to the communication number supervision end and the communication number receiving end, and the fraudulent number is accurately identified by the dual-end early warning linkage, including: sending the accurate user portrait of the fraudulent number to the communication number supervision end and the communication number receiving end; at the communication number receiving end, the fraudulent early warning information is first sent, and the communication number receiving end is warned that its communication method has been obtained by the fraudulent number end, and a dual-end linkage early warning is carried out with the communication number supervision end, the communication method leakage path is traced, and the fraudulent number is accurately identified to obtain the fraudulent sensitive information channel, and the fraudulent number is accurately identified by dual-end early warning linkage tracing;

[0115] Intelligent determination of the fraud direction of sensitive information involving fraud includes: classifying existing fraud types through information identification based on sensitive information involving fraud, and identifying and predicting new fraud types; determining the first decision path of the fraud direction in the first decision path based on the existing fraud type, and if the first decision path determines that it is an existing fraud type, directly performing a user portrait of the existing fraud type of the fraud number involved; if the first decision path determines that it is not an existing fraud type, performing a determination through the second decision path until a new fraud type is determined to be identified, predicting the development trend of the fraud direction of the new fraud type, and performing a user portrait of the new fraud type and a portrait of the development trend of the new fraud type; identifying and predicting the direction of the new fraud type based on the user portrait of the new fraud type and the portrait of the development trend of the new fraud type; intelligent determination of the fraud direction includes: determination of the financial fraud direction, determination of the asset fraud direction, and determination of the abduction fraud direction. Determination or gambling fraud direction determination; financial fraud direction determination includes: when the sensitive information includes fund fraud sensitive information, transfer fraud sensitive information or loan fraud sensitive information, it is determined to be financial fraud direction; asset fraud direction determination includes: when the sensitive information includes real estate fraud sensitive information, company fraud sensitive information or mortgage fraud sensitive information, it is determined to be asset fraud direction; abduction fraud direction determination includes: when the sensitive information includes address fraud sensitive information, transportation fraud sensitive information or travel fraud sensitive information, it is determined to be abduction fraud direction; gambling fraud direction determination includes: when the sensitive information includes gambling fraud sensitive information, winning fraud sensitive information or gambling fraud sensitive information, it is determined to be gambling fraud direction; intelligent fraud direction determination is performed based on the existing fraud type directions and predicted new fraud type directions.

[0116] The principle and effect of the above technical solution are as follows: the fraud-related direction tracing and identification subsystem includes: the fraud-related direction intelligent judgment subsystem, which makes intelligent judgment on the fraud-related direction of sensitive information according to the information verification result of the expert model of sensitive information involved in fraud; the fraud-related number precise user portrait subsystem, which makes precise user portrait of fraud-related numbers and accurately identifies fraud-related numbers through intelligent judgment on the fraud-related direction; the dual-end early warning tracing and identification subsystem, which sends the precise user portrait of fraud-related numbers to the communication number supervision end and the communication number receiving end, and conducts dual-end early warning linkage tracing and accurate identification of fraud-related numbers; sends the precise user portrait of fraud-related numbers to the communication number supervision end and the communication number receiving end, and conducts dual-end early warning linkage and precise identification of fraud-related numbers, including: sending the precise user portrait of fraud-related numbers to the communication number supervision end and the communication number receiving end, and conducting dual-end early warning linkage and accurate identification of fraud-related numbers The communication number monitoring end and the communication number receiving end are sent to the communication number receiving end; at the communication number receiving end, a fraud warning message is first sent, and the communication number receiving end is warned that its communication method has been obtained by the fraud number end, and a two-end linkage warning is carried out with the communication number monitoring end to trace the leakage path of the communication method, accurately identify the fraud number to obtain the fraud-related sensitive information channel, and conduct a two-end warning linkage tracing to accurately identify the fraud-related number; the fraud-related sensitive information is intelligently determined in the fraud direction, including: according to the fraud-related sensitive information, the existing fraud type direction is classified through information identification, and the new fraud type direction is identified and predicted; according to the existing fraud type, the first decision path determination of the fraud-related direction is carried out in the first decision path. If the first decision path is determined to be an existing fraud type, the fraud-related direction is directly determined. User portraits of existing fraud types of the number; if the first decision path determines that it is not an existing fraud type, the second decision path is used for judgment until a new type of fraud is identified, the fraud direction development trend of the new type of fraud is predicted, and user portraits of the new type of fraud and prediction portraits of the direction development trend of the new type of fraud are made; based on the user portraits of the new type of fraud and the prediction portraits of the direction development trend of the new type of fraud, the direction of the new type of fraud is identified and predicted; intelligent judgment of the direction of fraud includes: judgment of the direction of financial fraud, judgment of the direction of asset fraud, judgment of the direction of abduction fraud or judgment of the direction of gambling fraud; judgment of the direction of financial fraud includes: when the sensitive information of fraud contains sensitive information of fund fraud, sensitive information of transfer fraud or loan fraud When sensitive information related to fraud is included, it is determined to be financial fraud; asset fraud direction determination includes: when the sensitive information related to fraud contains sensitive information related to real estate, company or mortgage, it is determined to be asset fraud; abduction fraud direction determination includes: when the sensitive information related to fraud contains sensitive information related to address, traffic or travel, it is determined to be abduction fraud; gambling fraud direction determination includes: when the sensitive information related to fraud contains sensitive information related to gambling, prize-winning or gambling, it is determined to be gambling fraud; intelligent fraud direction determination is performed based on existing fraud types and predicted new fraud types;It can accurately identify fraudulent numbers through accurate user profiling of communication numbers and early warning linkage tracing, greatly increasing the accuracy of fraudulent number identification, significantly improving the timeliness of fraudulent early warnings, and significantly improving the efficiency of fraud tracing by timely tracing the source of fraudulent information channels.

[0117] Although the embodiments of the present invention have been disclosed as above, they are not limited to the applications listed in the specification and the implementation methods. They can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily implemented. Therefore, without departing from the general concept defined by the claims and the scope of equivalents, the present invention is not limited to the specific details and the illustrations shown and described herein.

Claims

1. Fraudulent number identification method based on user portrait, It is characterized in that include: S100, based on the communication number historical record big data, perform preliminary user profiling on the communication number, and obtain preliminary user profiling data of the communication number; S200, based on the preliminary user portrait data of the communication number, performing a portrait data fraud analysis and prediction through a portrait data analysis and prediction model to obtain the portrait data fraud analysis and prediction information; S300, based on the fraud analysis prediction information of the portrait data, verify the information through the expert model of fraud-related sensitive information, and obtain the information verification result of the expert model of fraud-related sensitive information; S400, based on the verification results of the expert model of sensitive information related to fraud, makes an intelligent judgment on the direction of fraud, and makes an accurate user portrait of the communication number. It sends the portrait to the communication number supervision end and the communication number receiving end, and conducts dual-end early warning linkage tracing to accurately identify the fraudulent number.

2. The method for identifying fraudulent numbers based on user portraits as claimed in claim 1, It is characterized in that S100 includes: S101, extracting communication number history record user features based on communication number history record big data; S102, creating a communication number history restoration service scenario by restoring the communication number service scenario according to the communication number history record user characteristics; S103, restore the business scenario according to the communication number history, create the communication number historical behavior pattern through behavior pattern analysis, make a preliminary user portrait of the communication number, and obtain the preliminary user portrait data of the communication number; the communication number historical behavior pattern analysis includes: communication number historical call record analysis, communication number historical SMS record analysis and communication number historical network interaction data analysis.

3. The method for identifying fraudulent numbers based on user portraits as claimed in claim 1, It is characterized in that S200 includes: S201, creating a profile data analysis and prediction model based on the preliminary user profile data of the communication number; performing fraud analysis and prediction on the profile data through the profile data analysis and prediction model, and analyzing and predicting whether the preliminary user profile data has a risk of fraud; S202, analyzing and predicting preliminary user portrait data output results according to the portrait data analysis and prediction model, and obtaining the portrait data fraud analysis and prediction information.

4. The method for identifying fraudulent numbers based on user portraits as claimed in claim 1, It is characterized in that S300 includes: S301, based on the collection of fraud-related sensitive information, an expert model of fraud-related sensitive information is created through intensive training of a fraud-related sensitive information integrated decision tree; S302, verifying the fraud-related sensitive information of the image data fraud-related analysis and prediction information through the fraud-related sensitive information expert model; S303, based on the verification of the sensitive information related to fraud, the results having the sensitive information related to fraud after verification are screened out, and the verification result of the expert model information of the sensitive information related to fraud is obtained.

5. The method for identifying fraudulent numbers based on user portraits as claimed in claim 1, It is characterized in that S400 includes: S401, based on the information verification result of the expert model of the fraud-related sensitive information, intelligently determine the fraud-related direction of the fraud-related sensitive information; S402, through intelligent determination of the fraudulent direction, accurate user profiling of the fraudulent number is performed, and the fraudulent number is accurately identified; S403, sending the accurate user portrait of the fraudulent number to the communication number supervision terminal and the communication number receiving terminal, and performing double-end early warning linkage tracing and accurate identification of the fraudulent number; The precise user portrait of the fraudulent number is sent to the communication number supervision end and the communication number receiving end, and the fraudulent number is accurately identified through dual-end early warning linkage, including: sending the precise user portrait of the fraudulent number to the communication number supervision end and the communication number receiving end; at the communication number receiving end, first sending fraudulent early warning information, and alerting the communication number receiving end that its communication method has been obtained by the fraudulent number end, and conducting dual-end linkage early warning with the communication number supervision end, tracing the communication method leakage path, accurately identifying the fraudulent number to obtain sensitive fraud information channels, and conducting dual-end early warning linkage tracing to accurately identify the fraudulent number.

6. Fraudulent number identification system based on user portrait, It is characterized in that include: The big data preliminary portrait subsystem performs preliminary user portraits of communication numbers based on the big data of communication number history records, and obtains preliminary user portrait data of communication numbers; The portrait data fraud analysis and prediction subsystem performs portrait data fraud analysis and prediction based on the preliminary user portrait data of the communication number through the portrait data analysis and prediction model to obtain the portrait data fraud analysis and prediction information; The fraud-related information expert model verification subsystem predicts information based on fraud-related analysis of portrait data, verifies information through the fraud-related sensitive information expert model, and obtains the information verification results of the fraud-related sensitive information expert model; The fraud-related direction tracing and identification subsystem, based on the information verification results of the expert model of fraud-related sensitive information, makes an intelligent judgment on the fraud-related direction, and makes an accurate user profile of the communication number. The profile is sent to the communication number supervision end and the communication number receiving end, and a two-end early warning linkage tracing is performed to accurately identify the fraud-related number.

7. The fraudulent number identification system based on user portrait as claimed in claim 6, It is characterized in that The big data preliminary portrait subsystem includes: User feature extraction subsystem, extracting user features of communication number history records based on communication number history record big data; The number business scenario restoration subsystem creates a communication number history restoration business scenario through communication number business scenario restoration based on the communication number history record user characteristics; The behavioral analysis preliminary profiling subsystem restores the business scenario based on the communication number history, creates the communication number historical behavior pattern through behavioral pattern analysis, makes a preliminary user profiling of the communication number, and obtains the preliminary user profiling data of the communication number; the communication number historical behavior pattern analysis includes: communication number historical call record analysis, communication number historical SMS record analysis and communication number historical network interaction data analysis.

8. The fraudulent number identification system based on user portrait as claimed in claim 6, It is characterized in that The fraud analysis and prediction subsystem of image data includes: The prediction model fraud analysis subsystem creates a profile data analysis prediction model based on the preliminary user profile data of the communication number; uses the profile data analysis prediction model to perform profile data fraud analysis and prediction, and analyzes and predicts whether the preliminary user profile data has fraud risks; The fraud prediction information output subsystem analyzes and predicts the preliminary user portrait data output results based on the portrait data analysis and prediction model, and obtains the portrait data fraud analysis and prediction information.

9. The fraudulent number identification system based on user portrait as claimed in claim 6, It is characterized in that The fraud information expert model verification subsystem includes: The fraud-related sensitive information expert model subsystem creates a fraud-related sensitive information expert model through fraud-related sensitive information integrated decision tree intensive training based on the collection of fraud-related sensitive information; The fraud-related sensitive information verification subsystem verifies the fraud-related sensitive information of the image data fraud analysis and prediction information through the fraud-related sensitive information expert model; The information verification result screening subsystem verifies the sensitive information related to fraud, screens out the results with sensitive information related to fraud after verification, and obtains the information verification results of the expert model of sensitive information related to fraud.

10. The fraudulent number identification system based on user portrait as claimed in claim 6, It is characterized in that The fraud-related direction tracing and identification subsystem includes: The fraud-related direction intelligent judgment subsystem makes an intelligent judgment on the fraud-related direction of the sensitive information based on the information verification results of the expert model of the sensitive information involved in fraud; The subsystem for accurate user profiling of fraudulent numbers can accurately identify fraudulent numbers through intelligent judgment of the fraudulent direction and accurate user profiling of fraudulent numbers; The dual-end early warning tracing and identification subsystem sends the accurate user portrait of the fraudulent number to the communication number supervision end and the communication number receiving end, and conducts dual-end early warning linkage tracing and accurate identification of the fraudulent number; The precise user portrait of the fraudulent number is sent to the communication number supervision end and the communication number receiving end, and the fraudulent number is accurately identified through dual-end early warning linkage, including: sending the precise user portrait of the fraudulent number to the communication number supervision end and the communication number receiving end; at the communication number receiving end, first sending fraudulent early warning information, and alerting the communication number receiving end that its communication method has been obtained by the fraudulent number end, and conducting dual-end linkage early warning with the communication number supervision end, tracing the communication method leakage path, accurately identifying the fraudulent number to obtain sensitive fraud information channels, and conducting dual-end early warning linkage tracing to accurately identify the fraudulent number.

Citation Information

Cited By

  • Anti-fraud full-link intelligent regulation and control method and device based on AI

    CN121980435A