TWC: Medium: Apollo: An Architecture for Scalable Verifiable Computing

云计算使计算能力有限的各方能够将大量数据的存储和计算外包。然而，由于实现错误或恶意行为，这种外包计算可能会被错误地执行。用于可验证计算的协议允许执行此类计算的不可信服务器也提供返回结果正确的简洁证明。利用密码学、编程语言、数据库和系统的专业知识，该项目开发了一种新的架构（Apollo），用于自动生成可扩展的可验证计算协议。Apollo有可能将可验证计算的速度提高几个数量级，同时扩展到大型数据集。本研究的重点是适用于不同设置的可验证计算的加密“小工具”，以及易于用这些低级小工具表达可验证计算的编程语言和类型系统。该项目的一个主要目标是开发一个可验证的SQL数据库的工作原型。 这项研究的结果将通过出版物和开源实施向广大受众传播，并将纳入马里兰州大学的云计算安全研究生课程。

Cloud computing enables computationally limited parties to outsource the storage of, and computation on, massive amounts of data. Such outsourced computations, however, might be performed incorrectly, due to implementation errors or malicious behavior. Protocols for verifiable computation allow an untrusted server performing such computations to also provide succinct proofs that the returned results are correct. Drawing on expertise from cryptography, programming languages, databases, and systems, this project develops a novel architecture (Apollo) for automatically producing scalable verifiable-computation protocols. Apollo has the potential to speed up verifiable computations by orders of magnitude while at the same time scaling to large datasets.This research focuses on cryptographic "gadgets" suited for verifiable computation in different settings, as well as programming languages and type systems for easily expressing verifiable computation in terms of these lower-level gadgets. A major goal of the project is to develop a working prototype of a verifiable SQL database. Results of this research will be disseminated to a wide audience through publications and open-source implementations, and will be integrated into a graduate course on cloud computing security at the University of Maryland.