Research On Privacy Protection Mechanism Of Android Application Based On Dynamic Security Strategy

In recent years,consumers have been exposed to the security risks of compromising their personal privacy while using mobile smartphones for daily work and entertainment activities.As the mobile operating system with the highest market share,Android is also subject to the most malicious attacks and security threats.Some Android applications demand far more system permissions than what are necessary for their functionality,and this abuse of the permission mechanism comes not only from malicious applications developed by attackers,but also from so-called benign applications developed by legitimate vendors,thus cannot be addressed by existing malicious detection methods.In addition,the definition and intensity of privacy protection often depends on specific requirements and usage scenarios,and the security model provided by the Android operating system based on system permissions cannot meet users' variable personal privacy protection needs.Another dilemma faced by users is that when they refuse to provide the necessary system permissions or personal information,the normal functioning of applications will be greatly deteriorated.In response to the above issues,we propose an Android application privacy protection mechanism which can dynamically restrict and proactively protect applications' dangerous permission access and sensitive interface invocation behavior according to users' security requirements and usage scenarios.In this thesis,we first propose an Android application permission access control method based on virtual machine bytecode injection technique,which can generate security strategy in the form of virtual machine bytecode and inject it into the code unit of applications involving dangerous permission requests and sensitive data access.The approach enables dynamic application behavior control and privacy protection by adjusting security strategies in real time while the applications are running.In order to ensure that users' restrictions on system permissions and sensitive interfaces would not affect the normal functionality of the applications and to protect users' personal privacy,we propose a differential-privacy-based application data desensitization method.The method desensitizes all the data streams shared by Android applications and servers,and is able to sever the association between data and personal information without affecting the normal data collection and analysis functions of applications,thus balancing the availability and security of the applications.The main work and innovations of this thesis are as follows.1.A security strategy injection and permission control method based on Android virtual machine bytecode is proposed.The method is capable of generating instantiated security strategies based on user-specific security requirements and injecting the security strategies in the form of virtual machine bytecode into the code units of Android applications involving dangerous permission requests and sensitive data access to achieve effective control and proactive protection against dangerous behavior of Android applications.2.To address the problem that existing bytecode injection techniques cannot dynamically adjust the control level and shield granularity according to changes in security requirements,and that bytecode needs to be reinjected when security strategies are updated,a dynamic security strategy adjustment mechanism based on Java reflection technique is proposed.This mechanism dynamically switches the level of permission control when security requirements change without having to reinject bytecode or repackage the applications.3.To address the problem that limiting necessary system permissions and sensitive calls would affect the normal functionality of applications,a privacy protection and data desensitization method based on differential privacy that protects the availability of applications while ensuring the security of private data is proposed.This method is based on a provable privacy guarantee for differential privacy and combines differential privacy with Gaussian Process through a disguised approach that can balance availability and security by ensuring the correctness of the statistical characteristics of the applied macro data while avoiding leakage of individual user privacy information.4.Based on the above-mentioned schemes and techniques,a prototype Android application privacy protection system is designed and implemented,which is capable of realtime dynamic control of applications' dangerous behavior at the bytecode level according to users' security requirements,and perform differential privacy desensitization of the shared data streams with servers.Evaluation shows that the security strategies injected into Android applications would not affect the application robustness in 85.27% of cases,on the basis of which it is able to effectively restrict 85.33% of sensitive API invocations and dangerous permission requests.Meanwhile,the impact of injected security strategies on the startup time of applications is at best only 0.635 s on average,increasing the size by about 0.372% on average after applications are repackaged.