移动平台的数据保护关系社会民生和国家网络空间主权。现有工作大多针对通过规整化函数接口获得的系统设备类敏感数据，并且基于这些规整化的接口进行数据保护。然而，海量移动应用软件中同样存在着价值巨大的用户私有敏感数据，并且这类敏感数据无法通过规整的函数接口获得，导致现有工作无法适用。于此同时，移动平台的新特性为用户私有敏感数据的保护研究带来了独特的挑战。针对用户私有敏感数据非规整的特点，本项目首先基于移动应用软件与服务器的交互模型给出非规整用户私有敏感数据的系统化定义，并且结合程序分析技术设计用户私有敏感数据的自动化识别方法。考虑到来自程序内外部的威胁，本项将从多个角度研究用户私有敏感数据的保护方法，包括防止用户私有敏感数据输出到公共可访问空间，防止用户私有敏感数据的输入界面被劫持，防止用户私有敏感数据被程序内部的第三方代码组件窃取。本项目的研究成果将显著促进移动网络空间数据保护理论和技术的发展。

Protection of sensitive data on mobile platforms is critical to the social stability and the national sovereignty in cyberspace. Existing works on data protection mainly focus on the sensitive data of operating systems and mobile devices which are acquired with structured function interfaces, and the protection is also built upon the structured interfaces. As the development of mobile computing these years, there are a huge number of sensitive data with high value in mobile applications, which we call user-centric sensitive data. Different to the sensitive data of operating systems and mobile devices, user-centric sensitive data could not be acquired with structured function interfaces, thus existing work could not apply to this kind of unstructured data. Meanwhile, the programming model and other distinct features of mobile platforms introduce significant research challenges in protecting user-centric sensitive data. To protect unstructured user-centric sensitive data, we first study the communication model between applications and servers to give a systematic definition of what is the user-centric sensitive data. Then, based on user-interface analysis and application behavior analysis, we design automatic techniques to identify user-centric sensitive data from applications. The protection of user-centric sensitive data is designed from two perspectives, i.e. protecting against inside and outside attackers of the application. Specifically, we detect insecure flows of user-centric sensitive to public accessible space to prevent information collection of outside attackers, monitor user interface spoofing attack to prevent outside attackers from hijacking user-input sensitive data, and design in-process data access management framework to prevent inner attackers from collecting user-centric sensitive data.