Research On Implicit Authentication Of Smart Phones Based On Context Awareness And Touch Dynamic

The wide application of smart phones has brought many conveniences to human life.At the same time,users are placing more demands on personal privacy protection in smart phones.The importance of authentication for smart phones is increasingly prominent.In recent years,as a new type of implicit authentication method,touch dynamic has gradually become a research hotspot in related fields due to its simplicity,lightweight,and reliability.Although the current research work on touch dynamic authentication has emerged in an endless stream,a large number of existing studies are conducted in the experimental environment.It means that many researchers only perform authentication analysis on users' specific touch behavior in a specific posture,without fully considering the impact of one's posture variability on the authentication performance in actual application scenario,which leads to a decline in the application performance of related technologies.Motivated by this,a series of solutions for enhancing the robustness and accuracy of touch-based authentication is proposed in this thesis.The main contributions of this thesis are as follows:(1)A context-aware touch dynamic authentication framework is designed to enhance the robustness of authentication.First,the proposed framework takes one's body posture in the process of authentication as the context information.Then,a context awareness module is introduced to enable users' touch behavior under different postures to achieve adaptive authentication by triggering a trained authentication model under the corresponding contexts.The experimental results show that compared to the traditional authentication framework,the proposed context-aware one can effectively reduce the impact of one's posture changes on authentication performance.(2)A user-independent context awareness algorithm is presented to distinguish users' various operating environments.Our work not only determines the minimum number of contexts that can effectively improve authentication performance,but also verifies the availability of context-aware algorithm by experiment analysis.(3)We propose a comprehensive feature set integrating gesture and touch dynamic characteristics by fusing multi sensor data.Specifically,it is the first time to propose and define "gesture feature".The touch-related features are obtained by in-depth analysis and optimization of common features.Our feature analysis results fully prove the superiority of the proposed feature set.(4)Based on the OC-SVM algorithm,a multi-level fusion decision scheme is designed to improve the accuracy of authentication.On the one hand,we put forward a kernel function based on feature contribution to achieve a weighted fusion at the feature layer;On the other hand,we raise a weighted metric-level fusion algorithm at the decision layer.Experimental simulations verify the effectiveness of the scheme.We collected a total of 10000+ samples from 87 users for experimental simulation.The experimental results show that: in the real application scenario,the average result of our proposed authentication scheme can reach a FAR of 0.7763% and FRR of 0.0005%for different passwords.Even for the simplest password "1111",the FAR obtained by authentication can reach 1.5386%,and the FRR can reach 0.Compared with the "European Access Control Standard"(FAR ? 0.001%,FRR ? 1%),our solution shows its high potential application value.