Social network topology model and construction method, user confidence and intimacy calculation method and telecommunication fraud intelligent interception system

Abstract

The invention discloses a social network topology model, and the model is a three-layer two-order star-shaped structure constructed by taking each target user as a center and comprises a first-layer node, second-layer nodes and third-layer nodes; the first-layer node is a target user, the second-layer nodes are a direct called user, and the third-layer node is obtained by modularization of neighbor nodes generated by each second-layer node; the node value of the social network topology model is the user confidence, and the edge weight of the social network topology model is the intimacy between users. The invention further discloses a construction method of the social network topology model, a user confidence coefficient and intimacy calculation method and a telecommunication fraud intelligent interception system. The social network topology model does not infinitely increase the data volume over time. According to the telecommunication fraud intelligent interception system, based on user historical data, the infinite increase of the system data size along with time can be avoided, and original and newly-added fraud users can be accurately identified.

Description

technical field [0001] The invention relates to the field of telecommunications, in particular to a social network topology model based on telecommunications call records. The present invention also relates to a construction method of the social network topology model, a user confidence iterative update subsystem utilizing the social network topology model, and a telecommunications fraud system with the user confidence iterative update subsystem Intelligent interception system. Background technique [0002] With the continuous development of society, the updating of science and technology and the wide application of Internet technology, telecommunications fraud has become one of the most common means of fraud, which has attracted widespread attention from all sectors of society. At present, various operators are actively studying various methods to deal with the problem of telecom fraud, so as to reduce economic losses. At present, a telecommunications fraud identification...

AI Technical Summary

Problems solved by technology

The advantage of this technology is that it can flexibly detect fraudulent users based on user behavior. The disadvantage is that it does not take into account the long-term social relationship between users and the trust degree of user historical behavior. When the number of ordinary users far exceeds that of fraudulent users, The performance of the model will be greatly limited

[0005] Chinese patent (CN201710721594.9) discloses a telecom fraud recognition and defense system based on deep learning. Its technical defect is that the existing telecom fraud recognition system uses speech recognition technology to detect abnormal information in the call content, which is This method requires the mobile phone to install the specified App, which will record all the call content of the user and convert it into text and upload it to the server

In terms of fraud identification technology, only the text content information in the current call is considered, and the historical behavior of the calling number, the intimacy relationship with the called number, etc. are not considered comprehensively, so it may cause misjudgment

[0006] Chinese patent (CN201910110374.1) discloses a communication fraud identification method and system, and its technical defects are: the existing fraudulent user identification technology is based on user call frequency, terminal location and whether the called party belongs to a group prone to fraud , to identify whether the calling party is communication fraud

This identification technology has relatively large limitations. In terms of judging the frequency of calls, it only uses a predetermined threshold for screening, and cannot reasonably distinguish fraudulent users, harassing users, and ordinary users with a large number of calls.

In terms of terminal location, it is susceptible to the influence of new fraudulent methods. If it is difficult for fraudulent users to distinguish from ordinary users, it is impossible to screen out fraudulent users.

In terms of judging whether the called number belongs to a group that is prone to fraud, it is impossible to judge the called user whose basic information cannot be mastered by the operator, which is not practical and affects the prediction of fraudulent users

[0007] Chinese patent (CN201610872103.6) discloses a device and method for preventing telecommunication fraud, and its technical defect is: the implementation of the existing method for preventing telecommunication fraud is based on the analysis of received voice and text information, It is necessary to conduct training and analysis for the specific call content of each bill

This method requires a detailed analysis of the content of each call record, and the amount of calculation is very large, so it is not suitable for processing massive call data

In addition, due to the protection of user privacy, it is difficult to obtain the specific content of the user's call or the content of the information sent between users, so in the actual application process, there are relatively large limitations

[0008] Chinese patent (CN201710211148.3) discloses a system and method for preventing telecom fraud based on big data and machine learning. Reporting, a fraudulent user being reported by many people often means that he has obtained illegal income through fraudulent behavior. At this time, the economic loss caused by shutting down the number and blocking fraud is irreparable

In addition, relying on machine learning models to conduct statistical analysis on text, language, and behavior separately, in the case of processing massive amounts of data, there will be a problem that the time required for data processing is too long

Images