Ford, Bryan AlexanderAlp, Enis Ceyhun2024-01-222024-01-222024-01-22202410.5075/epfl-thesis-8858https://infoscience.epfl.ch/handle/20.500.14299/203077Smart contracts have emerged as the most promising foundations for applications of the blockchain technology. Even though smart contracts are expected to serve as the backbone of the next-generation web, they have several limitations that hinder their widespread adoption, namely limited computational functionality, restricted programmability, and lack of data confidentiality. Moreover, addressing these challenges manually in application-specific ways requires a lot of developer effort and time due to the monolithic architecture of smart contracts. In this dissertation, we start over with a novel architecture for building and deploying general-purpose decentralized programs. To this end, we first propose a new architecture that replaces the monolithic execution model of smart contracts with a modular one to support a rich set of functionality, which can be easily and permissionlessly extended at any time. Second, to support the efficient deterministic execution required by computationally-advanced smart contracts, we build a deterministic sandbox with floating-point arithmetic support that brings safe and deterministic execution together with general-purpose programming without having to sacrifice performance. Finally, we combine threshold cryptography and the blockchain technology to build a framework that enables mutually distrustful parties to share their confidential data in a fully auditable, transparent and decentralized manner. Through prototyping and evaluation using real-world applications, we demonstrate that it is possible and feasibly-practical to build a decentralized computing platform that can support general-purpose computations.endecentralized computingsmart contractsdeterministic executionconfidentialitypost-quantumdata-flow graphsextensibilitymodularblockchaindata sharing.thesis::doctoral thesis