Writing and Securing Peer-to-Peer Computation
Abstract
This grant aims to make it easier to write and secure peer-to-peer computation: direct sharing of data and code between devices without a trusted server to mediate their interaction. The ability for peers to send data and computation directly to each other enables exciting applications in settings where a central trusted server is not available (due to either connectivity issues or insufficient trust in available servers).The objectives of this research lie in three primary directions. 1. Efficient enforcement of security and resource usage on code received from an untrusted device. 2. Efficient verification of code executed by an untrusted device. 3. Design and implementation of the Calder programming language, a language for sharing data and computation among heterogeneous resource-constrained devices. In this third year of the project, we continued the design and implementation of the Calder programming language. We have completed implementation of a simulator for execution of Calder programs and furthered development of Calder's runtime system to dynamically decide where to execute mobile code to optimize resource usage and to enforce resource constraints. Due to the departure of key personnel, we pivoted and spent significant effort this year on the use of trusted execution environments to enable efficient enforcement of security on code between peers. Specifically, we developed the DFLATE calculus, which provides high-level abstractions for using Trusted Execution Environments (such as Intel's SGX enclaves) to enable secure sharing of data and computation. This work was published at the IEEE Computer Security Foundations Symposium, a premier venue for foundational research on computer security.
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 15, 2019
- Accession Number
- AD1097237
Entities
People
- Stephen Chong
Organizations
- Harvard College