Research And Implementation Of User Authentication On Smartphone Using Keystroke Biometrics

Smartphones are widely used in daily life,and the mobile phone stores a large amount of social information,photo,voice,financial and other private information.After the mobile phone is stolen,once the privacy information is disclosed,it will pose a threat to the personal and property safety of the user.Therefore,the smartphone identity authentication system needs to be continuously enhanced as an important mobile phone intrusion detection measure.Considering the security and ease of use of mobile identity authentication,there are three main types of identity authentication methods.Knowledge-based authentication,such as account passwords,is easy to guess or crack.Token-based identity authentication,such as electronic secrets,is easy to lose and costly to carry.Identity authentication based on biophysical characteristics,such as fingerprint and faces,is a static identity authentication that is vulnerable to oil analysis or image forgery.Therefore,this paper studies a dynamic identity authentication method based on biological behavior characteristics,which does not bring the burden of memory or portability to the user,and is difficult to be copied.Based on the research of identity authentication knowledge,this paper conducts in-depth research and analysis on user mobile phone bio-behavior identity authentication.The prior researches extract the biological behavior characteristics of the user,calculates the correlation between the features and the categories,and selects the features with strong correlation to classify,and the error rate is relatively high.Aiming at this problem,this paper analyzes the biological behavior characteristics adopted by the current technology,defines the 146-dimensional behavior feature set including time feature,acceleration feature,pressure feature and size feature,and optimizes the feature combination by differential evolution algorithm.Combined with differential evolution feature optimization algorithm and support vector regression algorithm,the optimal error rate is about 0.12660%,and the battery consumption is reduced by about 31.25%.In addition,the existing methods mostly perform feature collection for a certain body posture of the user,while ignoring the fact that the user often uses the mobile phone in a multi-body posture,and the robustness of the model is not strong.In this paper,data collection is carried out for four kinds of body postures for legal user,such as station,sitting,lying,walking,etc.The joint quadrant range is extracted for each feature,and the anti-noise samples are constructed through the range,and then the model is tested and strengthened.Finally,in the strict experimental environment and noise environment,the error rate of the original method increased by about 165.63%,and the increase of this method was only 37.05%.This paper designs and implements a smart phone identity authentication system based on keystroke dynamics.The system includes an information acquisition module,an accuracy improvement module,a robustness test and enhancement module,and an identity verification test module.The information acquisition module completes the Android application development acquisition data;the accuracy improvement module completes the differential evolution plus support vector regression model construction;the robustness test and enhancement module completes the anti-noise sample construction and model retraining enhancement;the identity authentication test module completes the test from an actual unknown user.Experimental results show that the system can effectively authenticate users in a variety of practical scenarios.