Secure computation provides cryptographic protocols that support collaborative applications with private inputs and outputs. We examine in this paper a collection of protocols that evaluate polynomials with private inputs and outputs using secure floating-point arithmetic. A first goal is to provides a comparative analysis of their tradeoffs and performance. Moreover, the analysis is an opportunity to discuss several general problems: the tradeoffs between complexity metrics, their effects on protocol performance, and the importance of the execution environment in the experimental evaluation of secure computation protocols.

Keywords: Cryptographic protocols, Secure multiparty computation, Secure floating-point arithmetic, Polynomial evaluation