"Dr. OS: Operating system analysis, debugging and security tools based on dynamic binary translation"

计算机操作系统(OS)为用户应用程序提供重要服务，包括控制物理计算机硬件的共享使用，同时隐藏物理设备访问的复杂性，并保护应用程序(以及延伸到用户)免受其他应用程序的不想要的访问，无论是意外的还是恶意的。此外，他们必须高效地提供这些服务，并不断适应必须管理的不断变化的物理硬件和新类别应用程序不断变化的需求。就像任何大型软件项目一样，操作系统也包含错误，这并不令人惊讶。然而，由于操作系统的特权地位，操作系统中的错误可能会威胁到委托给它的所有应用程序和用户数据的可靠性和安全性。更令人惊讶的是，与广泛用于提高用户级软件应用程序质量的程序开发工具相比，OS开发人员可用的程序开发工具的状态很差。 动态二进制翻译(DBT)是一种强大的技术，它需要监控并潜在地在现有程序二进制中的每条指令执行之前对其进行操作。在过去，它被广泛用于为用户级应用程序软件开发各种分析、错误查找和安全工具。这些工具的使用极大地提高了它们所应用的应用程序的性能、健壮性和安全性。不幸的是，这些工具目前还不能用于操作系统，因为操作系统内核没有全面的DBT框架。由于内核代码的大小和复杂性不断增长，其正确性对整个计算机系统的整体健壮性和安全性的重要性，以及尽管静态错误检测工具取得了进展，内核错误仍在继续流行，我们最近开发了一个内核内DBT框架。我们的研究将提高该框架的性能，并开发用于内核分析、调试、安全审计和安全策略执行的工具。提高支持所有其他用户应用程序的操作系统的可靠性和安全性，将提高计算机系统的整体可信性。

Computer operating systems (OSs) provide vital services to user applications, including controlling the shared use of the physical computer hardware while hiding the complexity of physical device access and protecting applications (and by extention, the users) from unwanted access by other applications, whether accidental or malicious. Moreover, they must provide these services efficiently and constantly adapt to shifting physical hardware that must be managed and the changing demands of new classes of applications. It is not surprising to find that, like any large software project, operating systems contain bugs. However, due to the OS's privileged position, bugs in the OS can threaten the reliability and security of all applications and user data that have been entrusted to it. A greater surprise is the poor state of program development tools that are available to OS developers, in comparison to those widely used to improve the quality of user-level software applications. Dynamic binary translation (DBT) is a powerful technique that entails monitoring and potentially manipulating every instruction in an existing program binary just before its execution. In the past, it has been employed extensively to develop various analysis, bug-finding, and security tools for user-level application software. The use of these tools has greatly improved the performance, robustness, and security of the applications to which they have been applied. Unfortunately, such tools are currently not available for the OS since no comprehensive DBT framework exists for the OS kernel. Motivated by the ever-growing size and complexity of kernel code, the importance of its correctness for the overall robustness and security of the entire computer system, and the continued prevalence of kernel bugs despite advances in static bug-detection tools, we have recently developed an in-kernel DBT framework. Our research will enhance the performance of this framework and develop tools for kernel analysis, debugging, security auditing, and enforcement of security policies. Increasing the reliability and security of the OS, which supports all other user applications, will improve the overall trustworthiness of computer systems.