Abnormal user detection method based on minimum risk of deep neural network

Abstract

The invention discloses an abnormal user detection method based on the minimum risk of a deep neural network. The method comprises firstly, preprocessing data of an abnormal user to obtain data of theabnormal user with the same amount of data of a normal user; secondly, using a Mini-batch gradient descent method to train a deep neural network and combining with the degree of harm of different abnormal users in an NOMA communication system to establish a decision table based on the minimum risk; and finally, introducing loss functions corresponding to the different decisions set by the decision table into the deep neural network to construct an abnormal user detection method based on the minimum risk of the neural network. The method solves the problem of NOMA system abnormal user detection based on big data through strong expression and analysis capabilities of the deep neural network to high-dimensional data, solves the problem of differences in the degree of harm of various abnormalusers in the NOMA system by introducing the minimum risk in the deep neural network, and provides new ideas and theoretical innovations for abnormal user detection.

Description

technical field [0001] The invention belongs to the field of abnormal detection, and in particular relates to an abnormal user detection method based on a minimum risk deep neural network. Background technique [0002] Due to the openness of wireless channels, with the development of wireless communication technology, there are more and more security problems. If there is no effective response strategy, it may cause immeasurable losses to wireless communication networks and legitimate users. Non-orthogonal multiple access (NOMA) technology has become one of the key technologies of the next generation mobile communication system (5G). With the development of NOMA, security issues in NOMA have also begun to receive attention and research. In the power domain NOMA, a corresponding power allocation scheme is derived based on the user's channel state information (Channel State Information, CSI), and superimposed information transmission is performed on the same frequency spectr...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to classify and detect abnormal users, propose a method for detecting abnormal users based on a deep neural network with minimum risk, realize the detection of abnormal users, and solve the problem that traditional abnormal user detection methods are difficult to achieve high monitoring efficiency and lack of Technical issues of different loss decisions for various abnormal users

Images