Research On Mobile Phone Identity Recognition Based On User Behavior Characteristics

With the increase in the possession of social smart phones and the rapid development of mobile-side information services,the security of sensitive data generated and accumulated on mobile phones has received increasing attention.At present,the identification schemes commonly used in mobile phones include digital passwords,pattern codes,fingerprint recognition,face recognition,and iris recognition.This kind of identity recognition is called one-time identification or static identification.Once static identification is passed,it is no longer concerned about whether the current mobile phone operator is the user himself or an intrusion user.In this condition,the private data stored in the mobile phone may leak.This text carries on the research to the identity problem of the intellectual mobile phone end,proposes two kinds of identification schemes based on the intellectual mobile phone user is in the process of inputting the authentication password and sliding the touch screen with personal unique behavior characteristic.(1)Identification based on digital password input features.The scheme is based on the user's identity recognition by the pressure on the touch screen,the contact area and the input time characteristics during the process of inputting the digital password.When using a digital password to log in to a mobile phone,it adds a layer of security protection to the original password identification,making it difficult for the intrusion user to obtain a real password.(2)Identification based on features of the touch slide gesture.The program uses the user's four kinds of sliding gesture features(including pressure,contact area,speed,gesture track length,time and other characteristics)to identify the user identity.In this paper,the OCSVM algorithm is applied to the authentication model,and the following experiments are performed based on the collected 28-character feature data:(1)Selecting the best kernel function of OCSVM algorithm through experiments,it is concluded that the RBF kernel performs better than the other three kernel functions(Linear,Polynomial,Sigmoid)in accuracy and model training time.(2)Comparing the advantages and disadvantages of the OCSVM classification algorithm used in this paper with other commonly used classification algorithms(SVM,BP neural network,Naive Bayes)training,the model of OCSVM algorithm trained in classification accuracy It is higher than other three classification algorithms and is second only to Naive Bayes algorithm in training time.(3)Analyze the impact of different test data volumes on model classification.Experiments have found that when the test data volume is higher than 300 and 800,the two FAR and FRR values are in a relatively stable state(identity based on password input characteristics).The FAR and FRR values were verified to be in the 0%-5.6% range,and the FAR and FRR values for the touch gesture-based authentication scheme ranged between 2.75% and 4.1%.(4)Comparing the experimental results of this paper with the experimental results of the related subjects in the past 5 years.The results show that the two identification schemes in this paper are lower than the experimental values of the two related literatures in the FAR values,showing that the Good recognition rate.In the performance of FRR,the two identification schemes in this paper are better than those in the literature [36],and are only inferior to the experimental results in [15].The experimental results of this topic show that the user's behavioral characteristics data can be used to effectively distinguish the user's identity.Two identity authentication schemes can improve the security of the device as an identity authentication mechanism.In practical applications,the operability is strong.The authentication process is transparent to the user,does not interfere with the user's normal operation,and does not require additional hardware overhead..Identity recognition based on digital password input features can be seamlessly integrated into existing cryptosystems,while identity recognition based on touch gesture features enables continuous monitoring and collection of gesture feature data generated by the user's finger interacting with the touchscreen.The continuous and dynamic identification of operating user identities ensures the security of private data in smart phones,and makes up for the inadequacies of the traditional authentication schemes that are only verified during the login phase.