DSCS: Towards an Extensible Secure Decentralised Distributed Computation Protocol

This paper documents DSCS, a privacy preserving distributed computation protocol with native support for homomorphic encryption. The protocol is easily extensible with advanced cryptographic constructions due to a program encoding with a bijective map to arithmetic circuits. Arithmetic circuits are a common mechanism for abstract computation, and are widely used by advanced cryptographic constructions, such as homomorphic encryption and verifiable computation. This paper documents foundational work exploring the validity of an arithmetic circuit based approach to distributed computation. Ultimately, the direct use of arithmetic circuits as a computational substrate for distributed computation is shown to carry an untenable bandwidth cost, even under ideal network conditions. Importantly, this paper highlights a slew of low hanging fruit for further research in this domain, as well as guidance to avoid easy pitfalls.

I. Introduction

The intention of this work is to define a secure-by-default decentralised and distributed computation protocol. Such a protocol would allow for mutually distrusting distributed devices, such as distributed sensors in a re-mote location, or Internet of Things (IoT) devices spread throughout a home, to collaborate on the execution of programs. Potentially, through collaboration with nearby devices the computational latency of individual task can be reduced, as devices leverage the spare computational capacity of their peers.