Use this URL to cite or link to this record in EThOS:
Title: Design and implementation of secure multi-party computation
Author: Scholl, Peter Alexander
ISNI:       0000 0004 5923 1524
Awarding Body: University of Bristol
Current Institution: University of Bristol
Date of Award: 2015
Availability of Full Text:
Access from EThOS:
In a secure multi-party computation (MPC) protocol, a number of parties jointly compute a function on their private inputs, learning only the output of the function. The SPDZ protocol is one of the most practical, actively secure MPC protocols tolerating a majority of corrupted parties. We describe a covertly secure distributed key generation protocol for the preprocessing stage of SPDZ, which was previously assumed as a global setup, and also present more efficient, covertly secure protocols for the remainder of the preprocessing. We then give a new approach to implementing the preprocessing using oblivious transfer, which provides an alternative foundation for this stage of the protocol. Using a novel actively secure protocol for extending oblivious transfers, we estimate that this should be much more efficient than existing protocols based on somewhat homomorphic encryption. Next, we turn to implementing the online phase of SPDZ, where the actual computation takes place. We design a virtual machine-based architecture for running the online phase and create a compiler that allows programmers without any experience of MPC protocols to easily implement algorithms in MPC, producing optimised code for the virtual machine. We present benchmarks of various common and useful functions - including secure comparison, floating point arithmetic and AES - running on our architecture. These are significantly faster than all previously published results with the same level of security. Finally, we present efficient, sec ure protocols for oblivious data structures such as arrays, priority queues and dictionaries using Oblivious RAM. As an application. of these, we show how to perform secure computation of Dijkstra's shortest paths algorithm on graphs, where the graph structure is secret. We give benchmarks for the resulting protocols using the previous architecture.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available