CAREER: Enhancing Mobile Application Security through Contextual Integrity and User Awareness

作为移动的应用（即，应用程序）已经成为日常生活的组成部分，它们对用户的敏感数据（例如，地点和联系人）引起了严重的安全关切。主流智能手机平台（例如，Android和iOS）采用基于权限的访问控制机制，但此类机制未能考虑权限请求产生的上下文，也没有解释应用程序如何以及为什么使用敏感数据，导致用户做出不知情的决定。该项目的目标是开发一个上下文和用户感知的安全框架，该框架通过仅在敏感数据以上下文和应用程序意图无法证明的方式使用时通知用户来实现（G1）上下文完整性，以及（G2）通过生成解释敏感数据使用的自然语言（NL）描述来实现用户感知。 研究将有四个主要任务。首先，研究团队将开发一个上下文和意图感知模型，该模型表示上下文/意图与代码中的敏感行为之间的相关性，其中上下文和意图主要由非结构化信息（即，图像和文本）。其次，该团队将开发新的程序分析技术，将GUI中的上下文和意图与代码中的敏感行为相关联，从而能够构建大规模的高质量训练数据。第三，该团队将开发一个神经机器翻译模型，该模型将GUI上下文提供的上下文信息和隐私策略提供的词汇作为输入，并合成代码中敏感行为的描述。最后，该团队将开发一个轻量级的仪器系统，该系统将检测到的不良行为的结果和合成的描述集成在一起。该项目的成功将通过产生更安全的移动的应用程序来增强整个社会的安全性，并且所提出的技术将为程序分析和机器学习的合作提供新的见解。该项目开发的新技术和工具将被整合到本科和研究生教育中，并用于提高公众对移动应用程序安全重要性的认识。该奖项反映了NSF的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

As mobile applications (i.e., apps) have become an integral part of daily life, their increasing access to users' sensitive data (e.g., location and contacts) raises serious security concerns. Mainstream smartphone platforms (e.g., Android and iOS) adopt permission-based access-control mechanisms, but such mechanisms fail to consider the context in which permission requests arise and do not explain how and why the app uses sensitive data, causing users to make uninformed decisions. The goal of this project is to develop a context- and user-aware security framework that enables (G1) contextual integrity by notifying users only when sensitive data is used in the ways that cannot be justified by the contexts and the apps’ intentions, and (G2) user awareness by generating natural-language (NL) descriptions that explain the sensitive data uses. The research will have four major tasks. First, the research team will develop a context- and intention-aware model that represents the correlation between the contexts/intentions and the sensitive behaviors in the code, where the contexts and intentions are expressed mainly by unstructured information (i.e., images and text) in the Graphical User Interfaces (GUIs). Second, the team will develop novel program-analysis techniques that associate the contexts and the intentions in GUIs to the sensitive behaviors in the code, which enables the construction of a large-scale high-quality training data. Third, the team will develop a neural machine-translation model that takes as input the contextual information provided by GUI contexts and the vocabulary provided by privacy policies, and synthesizes descriptions for sensitive behaviors in the code. Finally, the team will develop a lightweight instrumentation system that integrates the results of the detected undesired behaviors and the synthesized descriptions. The success of this project will enhance the security of society at large by leading to more secure mobile apps, and the proposed techniques will provide new insights for the cooperation of program analysis and machine learning. New techniques and tools developed in this project will be integrated into undergraduate and graduate education and used to raise public awareness of the importance of mobile-app security.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.