SHF:Small:Collaborative Research: Compositional Verification of Heterogeneous Software Protocol Stacks

开发软件协议堆栈是一项复杂的任务，涉及多个方面和产品多年。在这个过程中有许多问题必须解决，而且要让所有问题都正确适用于所有平台和系统的所有可能状态是令人难以置信的困难。因此，有许多严重错误的例子最终在生产中结束，修复它们的成本很高。软件协议栈在很多领域都有广泛的应用，采用严格的方法来提高其可靠性将产生非常广泛的影响。该项目开发了一种更严格的方法来开发和维护这种复杂的软件协议栈，特别是使用正式方法通过严格的验证和测试及早发现错误，并帮助分析定位报告的错误。该方法基于自动学习目标软件堆栈不同层抽象和简洁的协议模型。然后应用成分推理来有效和准确地找出错误。生成的模型被用来帮助错误定位和诊断。通过对Android蓝牙软件堆栈的真实案例研究，评估了该方法的可用性。

Developing a software protocol stack is a complex task with many parties and products involved over multiple years. There are many issues that have to be resolved in the process, and it is incredibly hard to get all of them correct for all platforms and all possible states of a system. Hence, there are numerous examples of serious errorsthat ended up in production, where the cost of fixing them is high. Software protocol stacks are widely used in many domains, and rigorous methods to improve their reliability would have a very broad impact. The project develops a more rigorous approach to developing and maintaining such complex software protocol stacks.In particular, formal methods are used to find errors early through rigorous verification and testing, as well as to assist the analysis in locating reported errors. The approach is based on automatically learning abstract and succinct protocol models at different layers of the target software stack. Compositional reasoning is then applied to find errors effectively and precisely. The generated models are leveraged to help with error localization and diagnosis. The usability of the approach is assessed through a real-life case study on the Android Bluetooth software stack.