Three-branch identity authentication method and system based on mouse behaviors

Abstract

The invention belongs to the field of identity authentication, and particularly relates to a three-branch identity authentication method and system based on mouse behaviors, and the method comprises the steps: carrying out mouse data collection, and carrying out duplicate removal processing and abnormal removal processing on the collected data; using a data sequence composed of a plurality of basic mouse operations as a mouse behavior, and a feature vector is calculated according to the mouse behavior, so that a related feature value is calculated; Modeling the obtained mouse features to obtain three identity authentication models, outputting probability values of samples belonging to legal mouse users by the models, obtaining threshold values required by classification through a loss function, and dividing sample objects into a positive domain, a boundary domain or a negative domain; and judging the negative domain sample object as an illegal mouse user, judging the positive domain sample object as a legal mouse user, and continuing to authenticate the boundary domain mouse user. According to the method, a decision is made in advance before the number or time of the mice is determined, and higher authentication precision is achieved within lower expected authentication time.

Description

technical field [0001] The invention belongs to the field of identity authentication, and in particular relates to a mouse behavior-based three-way identity authentication method and system. Background technique [0002] Achieving reliable identity authentication is a key issue in protecting the security of applications and services in the network, and the method based on mouse behavior characteristics does not require additional expensive hardware equipment, and only needs the mouse used by the present invention to collect the interactive data when the present invention uses a computer , extract useful features, and use these features to train a classification model to authenticate the legitimacy of the mouse user's identity. Many scholars have proposed an authentication method based on mouse behavior. In the "Identity Authentication Method Based on Mouse Behavior of Mouse Users" proposed by Xu Jian et al., a preliminary method for identifying legitimate mouse users based ...

AI Technical Summary

Problems solved by technology

[0003] In the existing methods, the binary classifier is directly trained, and the mouse user is directly judged as a legal mouse user or an illegal mouse user, which leads to the disadvantage that the existing model of identity authentication needs to correspond to a high authentication time in order to obtain high authentication accuracy.

However, some mouse users can judge the legality of their identity in advance, which will waste a lot of time

Images