Efficient Multiparty Computation for Comparator Networks

We propose a multiparty protocol for various computations using comparator networks such as sorting and searching. By repeating the execution of a comparator, the proposed protocol can efficiently detect outlier values, without revealing them. In our scenario, all input values to a comparator network and the intermediate output from each comparator are kept secret assuming the presence of an honest majority. Possible application areas for the proposed protocol include statistical analysis while preserving the privacy of respondents.