Parallel mixing
Philippe Golle, Ari Juels · 2004
Efforts to design faster synchronous mix networks have focused on reducing the computational cost of mixing per server. We propose a different approach: our reencryption mixnet allows servers to mix inputs in parallel. The result is a dramatic reduction in overall mixing time for moderate-to-large numbers of servers. As measured in the model we describe, for n inputs and $M$ servers our parallel re encryption mixnet produces output in time at most 2n -- and only around n assuming a majority of honest servers. In contrast, a traditional, sequential, synchronous re-encryption mixnet requires time Mn.