Authors:
Mohammad Anagreh
1
;
2
;
Eero Vainikko
2
and
Peeter Laud
1
Affiliations:
1
Cybernetica, Mäealuse 2/1, Tallinn, Estonia
;
2
Institute of Computer Science, University of Tartu, Narva maantee 18, Tartu, Estonia
Keyword(s):
Privacy-preserving Computation, Multi-party Computation (SMC), Single-Instruction-Multiple-Data (SIMD), Protocol of Reading Private Array, Prim’s Algorithm, Minimum Spanning Tree, Sharemind.
Abstract:
In this paper, we propose a secret sharing based secure multiparty computation (SMC) protocol for computing the minimum spanning trees in dense graphs. The challenges in the design of the protocol arise from the necessity to access memory according to private addresses, as well as from the need to reduce the round complexity. In our implementation, we use the single-instruction-multiple-data (SIMD) operations to reduce the round complexity of the SMC protocol; the SIMD instructions reduce the latency of the network among the three servers of the SMC platform. We present a state-of-the-art parallel privacy-preserving minimum spanning tree algorithm which is based on Prim’s algorithm for finding a minimum spanning tree (MST) in dense graphs. Performing permutation of the graph with sharemind to be able to perform the calculation of the MST on the shuffled graph outside the environment. We compare our protocol to the state of the art and find that its performance exceeds the existing pr
otocols when being applied to dense graphs.
(More)