Fraud number identification method and system based on convolutional neural network

Abstract

The invention discloses a fraud number identification method and an identification system based on a convolutional neural network. The identification method comprises the following steps: 1, establishing a training sample set: acquiring a plurality of numbers known as a customer service number, a private number and a fraud number, acquiring call data and M call features of each number for continuous N days, constructing an N * M feature matrix and converting the N * M feature matrix into a feature map, and forming a training sample by the feature map and a number category; 2, establishing a fraud number recognition model, and training the fraud number recognition model by adopting the training sample set; 3, obtaining call data and M call features of the telephone number to be identified for continuous N days, constructing an N * M feature matrix, and converting the N * M feature matrix into a feature map; and 4, performing classification identification on the feature map of the to-be-identified number by adopting the trained model to obtain a category label. According to the method, the distinguishing features and the combination of the distinguishing features of the fraud number,the customer service number and the common private number are extracted through deep learning, and the fraud number can be accurately recognized.

Description

technical field [0001] The invention belongs to the technical field of security identification of calling numbers in communications, and in particular relates to a method and system for identifying fraudulent numbers. Background technique [0002] With the popularization of mobile phones, phone scams emerge in endlessly. Although the relevant government departments have issued reminders to the society and various news media have reported frequently, however, a large number of users are still being cheated every day, and the economic losses are increasing year by year. [0003] In the prior art, the process of identifying and triggering fraudulent numbers is usually as follows: First, all calls in the entire network are triggered to the SCP, and then the SCP identifies a large number of calls one by one, and transfers a small number of suspected fraudulent calls to the anti-fraud platform. The anti-fraud platform will record and collect evidence of suspected fraudulent calls...

AI Technical Summary

Problems solved by technology

[0005] 1. The amount of call data is large, and the time cost required for calculation is high;

[0006] 2. The identification of fraudulent numbers is difficult: due to the increasingly "smart" criminal methods, and even the phenomenon of "pre-job training" and "on-the-job education", the characteristics of fraudulent numbers are complex, and it is difficult to obtain their in-depth characteristics with simple statistical analysis schemes;

Images