TWC: Small: Collaborative: Improving Android Security with Dynamic Slicing

移动的设备已经非常成功，并继续扩大其用户群。 然而，使这些设备成功的特性，例如，丰富的传感器输入（GPS、摄像头、麦克风）和持续的互联网连接也使这些设备成为攻击者最喜欢的目标。 攻击可能会产生许多负面后果，从窃取用户的秘密到监视用户或安装使设备无法运行的病毒。 该项目将开发动态切片技术，以便开发人员和研究人员能够有效地洞察设备和应用程序的行为，包括恶意应用程序;这将使构建安全应用程序和发现/消除恶意行为变得更加容易，这反过来又将使移动终端用户受益。本课程将向本科生和研究生介绍智能手机安全的新方法，这将使学生更好地应对新兴的软件研究和开发挑战。动态切片（分析执行以识别相关代码和数据依赖关系）是解决广泛安全问题的特别有效的技术。该项目将为Android开发一个动态切片器，然后使用切片器，以及与其他现有工具的集成，用于三个主要的安全应用程序。 首先，通过高效、有效、集成的控制/数据切片来改进动态污点分析。第二，在输入中找到相关部分以识别哪个传感器流及其部分负责安全相关行为，例如，攻击、崩溃、僵尸网络操作或使用反检测技术。第三，撤销计算，特别是将切片与记录和重放相结合，以支持Android上的撤销计算。 预计这些工作领域将在以下方面取得进展：精确和有效的动态污点跟踪、由于控制依赖性而发现泄漏、理解僵尸网络行为、反检测技术的原则性发现、将良性状态改变与恶意状态改变分开;以及程序分析，例如，对面向事件或严重依赖进程间通信的程序进行切片;在存在高吞吐量事件流的情况下计算输入/数据/代码相互依赖性。

Mobile devices have been very successful and continue to expand their user base. However, the very features that have made these devices successful, e.g., rich sensor inputs (GPS, camera, microphone) and continuous Internet connectivity, have also made the devices a favorite target for attackers. Attacks can have many negative consequences, from stealing users' secrets to spying on the users or installing viruses that render devices inoperable. This project will develop dynamic slicing techniques so that developers and researchers will be able to gain effective insights into device and app behavior, including malicious apps; this will make it easier to construct secure apps and to find/eliminate malicious behavior, which in turn will benefit mobile device users. Undergraduate and graduate students will be introduced to new approaches to smartphone security, which will make students better equipped for tackling emerging software research and development challenges.Dynamic slicing (analyzing an execution to identify relevant code and data dependences) is a particularly effective technique for addressing a wide range of security problems. This project will develop a dynamic slicer for Android and then use the slicer, as well as its integration with other existing tools, for three main security applications. First, improving dynamic taint analysis via efficient, effective, integrated control/data slicing. Second, finding relevant parts in the input to identity which sensor stream and part thereof are responsible for security-relevant behavior, e.g., attack, crash, botnet operation, or use of anti-detection techniques. Third, Undo Computing, in particular combining slicing with record-and-replay to support undo computing on Android. These lines of work are expected to lead to advances in: security, e.g., precise and effective dynamic taint tracking, finding leaks due to control dependences, understanding botnet behavior, principled discovery of anti-detection techniques, separating benign from malicious state changes; and program analysis, e.g., slicing programs that are event-oriented or rely heavily on inter-process communication; computing input/data/code interdependences in the presence of high-throughput event streams.