CNS Core: Small: Rethinking Runtime Software Security Hardening in the Context of Hybrid Instruction Set Architecture

计算的异构性越来越强(例如，CPU/GPU、CPU/FPGA)，以满足应用程序不断增长的性能需求，这在一定程度上是摩尔定律的终结。这一方向的最新趋势是具有不同指令集架构(或ISA)的CPU，特别是在移动计算和物联网(IoT)设置中。异类计算机的应用程序通常是通过将它们划分到ISA边界来开发的：由于固有的体系结构差异，每个ISA的软件都是单独开发的。因此，应用程序程序员通常不能充分利用特定于ISA的硬件安全功能。此外，现有的软件安全技术在很大程度上是特定于ISA的：威胁模型和利用缓解通常是为目标ISA开发的。该项目通过开发允许应用程序透明地利用特定于ISA的安全功能的软件基础设施来解决这一问题。该基础结构包括一个编译器和一个运行时框架，该框架通过为多个ISA生成代码、用于跨ISA安全强化的动态进程状态转换以及跨ISA统一共享内存来隐藏ISA差异并公开与ISA无关的抽象。将使用该基础设施开发几种安全技术，以展示其有效性，包括自适应攻击面缩减、代码和数据空间随机化以及动态软件修补。该项目的软件基础设施将使程序员能够为新兴计算机开发更安全的应用软件，而无需额外的编程工作，从而满足社会对安全计算基础设施的迫切需求。该项目的软件产品和实验结果将在项目完成后至少五年内以许可的开源许可证在项目网站上公开提供。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

Computing is increasingly heterogeneous (e.g., CPU/GPUs, CPU/FPGAs) to meet the ever-increasing performance demands of applications, necessitated in part by the end of Moore's law. The latest trend in this direction is CPUs with different instruction-set-architectures (or ISAs), especially in mobile computing and Internet-of-Things (IoT) settings. Applications for heterogeneous computers are often developed by dividing them across the ISA boundary: software for each ISA is separately developed due to the intrinsic architecture differences. As a result, application programmers often cannot fully leverage ISA-specific hardware security features. In addition, existing software security techniques are largely ISA-specific: threat models and exploit mitigations are usually developed for a target ISA. This project addresses this problem by developing a software infrastructure that allows applications to transparently utilize ISA-specific security features. The infrastructure includes a compiler and a run-time framework that hides ISA differences and exposes an ISA-agnostic abstraction through code generation for multiple ISAs, dynamic process state transformations for security hardening across ISAs, and cross-ISA unified shared memory. Several security techniques will be developed using this infrastructure to demonstrate effectiveness including adaptive attack surface reduction, code and data space randomization, and dynamic software patching. The project’s software infrastructure will allow programmers to develop more secure application software for emerging computers without additional programming effort, and thereby meet the pressing societal need for secure computing infrastructures. The project's software artifacts and experimental results will be publicly available under a permissive open-source license at the project website for at least five years after the project completion.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.