TWC: Medium: Scaling proof-based verifiable computation

这项研究解决了系统安全中的一个基本问题：一台机器如何将一项计算指定给另一台机器，然后在不执行计算的情况下检查另一台机器是否正确执行了该计算？在过去的几年里，出现了一种解决这个问题的新方法，它基于改进加密和理论工具，并将它们整合到构建的系统中。然而，尽管取得了令人兴奋的进展，但由此产生的系统在正常意义上仍然不实用。这个项目的目标是在实际的输入规模下对某些实际问题的真正实际性能。这项研究可能会对第三方计算模型产生广泛的影响，例如云计算、分布式系统、不可信的硬件制造商和国防应用程序。更广泛地说，真正廉价的可验证计算将使构建系统的新方法成为可能。该项目应用了编程语言社区的想法来更有效地表示计算；应用程序近似的概念来产生有效的部分可验证性；改进和实现承诺降低内存利用率的协议；并开发新的协议，减少开销，更适合重要的计算类别。方法包括改进和调整现有理论，在改进的基础上实现系统，建模性能和成本，并进行严格的实验评估。研究结果将通过同行评审的出版物以及免费提供的软件和实验配置进行传播。

This research addresses a fundamental problem in systems security: how can a machine specify a computation to another one and then, without executing the computation, check that the other machine carried it out correctly? Over the last several years, a new approach to this problem has emerged, based on refining cryptographic and theoretical tools, and incorporating them into built systems. However, despite exciting advances, the resulting systems are still not practical in the normal sense. This goal of this project is truly practical performance for certain real-world problems at realistic input sizes. This research could have wide-ranging impact on third-party computing models such as cloud computing, distributed systems, untrusted hardware manufacturers, and defense applications. More broadly, truly inexpensive verifiable computation will enable new ways of building systems. The project applies ideas from the programming languages community to represent computation more efficiently; applies notions of program approximation to produce efficient partial verifiability; refines and implements protocols that promise reduced memory utilization; and develops new protocols that reduce overhead and are better-suited to important classes of computations. The methods include refining and adapting existing theory, implementing systems based on the refinements, modeling performance and costs, and conducting rigorous experimental evaluations. The results will be disseminated through peer-reviewed publications, and through software and experimental configurations that will be available freely.