Research On Android Privacy-preserving Model With Application Behavior Monitoring

With the rapid development of mobile Internet,smartphones,one of the access terminal devices in the mobile internet,have been widely used around the world,and have become one of the most important tools for people's daily work and social entertainment.Android system is the most widely used mobile operating system in mobile smartphone systems because of its good user experience and open source.A strong technical support from Google and an influx of developers prompt a lot of the birth of the third party applications.However,malicious third-party applications may directly or indirectly exploit the system vulnerabilities to steal the users' sensitive information without the consent of users.The applications,spontaneously or conspiratorially with other applications,send the sensitive information to the remote server,leading to a serious leakage on users' privacy information.Therefore,the android privacy protection has become an urgent problem to be solved.To solve the problem,we designed a privacy protection model-AMDroid privacy model in this paper,which is mainly achieved through the code modification of Android framework layer.This model contains the three different modules,namely,the ApplicationDiscriminant model,the Permission-Management module and the Internet Start-up Management module.Our main contents are listed as follows:1.We implement the Application-Discriminant module based on machine learning methods.This module firstly learn the application permission of the training data,and then construct the classification model.Finally the installing applications can be classified as the malicious application and the normal application.The module can also present the classification results in a friendly way to users to help them understand whether the application being installed is harmful.2.We achieve the Permission-Management module based on the classification result of the Application Discriminant module.This module can set the dangerous application permissions and communication channel permissions for the malicious application to a dynamic authorization.We prevent the malicious application from stealing users' privacy data by defining the fine-grained application permissions and strengthening the access control.For the normal application,This module avoid too much authorization operation to enhance the users' experience by setting the applying application permissions to a normal permission uniformly.3.We realize the Internet Start-up Management module based on the classification result of the Application Discriminant module.We protect the users' privacy and prevent the collusion attack among applications by monitoring and controlling the interconnection between the malicious applications and other applications.Finally,we realize the AMDroid model based on the Android 6.0 version.In our experiment,we prove that our model can classify the installed applications as malicious and normal applications effectively,define the fine-grained or extensive permissions management,and prevent the interconnection among applications.Our system has a highlevel robustness.