2007.09.14 | Jesper Buus Nielsen

Secure multiparty computation allows n participants to compute an agreed
function securely, such that noinformation is revealed about the input of
the function,other than what is revealed by the function itself. This
thesis will present a Java based implementation of multiparty protocols with
communication complexity O(C*n)*k + O(D*n^2)*k + poly(n*k), where C is
the size of a circuit computing the desired function, n is the number of
parties, k is a security parameter and D is the depth of the circuit. The
theoretical foundation is briefly presented to the reader with references for
further details. The protocols are based on the idea of using a linear error
correcting code for avoiding active adversaries.

The design of the implementation is focused on generic use, without a spe-
cific target. However, a small demonstration applicationis presented in the
end of the thesis. This demonstrationapplication is only meant for
demonstration and all necessary certificates are bundled with the
application.

Theimplementation includes a small library for handling calculations in the
finite fields Z_p and Z_p[X]/poly(X), where p is some prime and poly(X) some
polynomial

The performance of the implementation has been evaluated and the results
corresponds to the theory. The evaluation highlights that the protocols main
workload lies in a so called preprocessing phase which can be conducted in
advance of the actual computation. The implementation supports that the user
saves this preprocessing result to a file and later loadsit when needed.
The computation itself is carried out quickly when the preprocessing has
been completed.