Identity recognition method based on abnormal behavior detection

Abstract

The invention discloses an identity recognition method based on abnormal behavior detection. The main technical scheme comprises the following steps of firstly, pre-embedding multiple behavioral habitmechanisms, then collecting behavior data based on a sensor and generating features by adopting a standardized preprocessing method; and according to the historical behavior characteristics of the password input by the target user, adopting an OCSVM algorithm training model to obtain the probability of behavior abnormality, combining results of multiple preset behavior habit mechanisms and finally, acquiring an identity authentication result in a linear combination mode. Only the target user sample exists in the training process, whether the new sample is the target user can be accurately andstably recognized, and therefore the judgment accuracy and stability can be improved.

Description

technical field [0001] The invention relates to the technical field of identification, in particular to an identification method based on abnormal behavior detection. Background technique [0002] With the development of the mobile Internet, smart phones have been widely used in people's lives. While the mobile Internet brings convenience, it is also accompanied by the threat of user information loss. Most APP applications on smart phones need to enter a password to identify whether they are operating in scenarios such as login and payment. However, only one password input operation is still very risky. When the password is stolen by an intruder through shoulder spying, etc. Finally, the intruder only needs to enter the correct password to steal all the information of the user's APP, and even transfer the money in the APP to other accounts. [0003] Although some APPs have already used fingerprint recognition, facial recognition and other methods for identification, these ...

AI Technical Summary

Problems solved by technology

[0007] The second is that we know that there are only data with normal behavior, so this situation belongs to the singular value detection problem of semi-supervised learning

[0013] The existing technology uses Z-score as a classification algorithm. Although it has achieved good results, the generalization ability of this technology is weaker than that of OCSVM.

Images