In order to guarantee a fair and transparent voting process, electronic voting schemes must be verifiable. Most of the time, however, it is important that elections also be anonymous. The notion of a verifiable shuffle describes how to satisfy both properties at the same time: ballots are submitted to a public bulletin board in encrypted form, verifiably shuffled by several mix servers (thus guaranteeing anonymity), and then verifiably decrypted by an appropriate threshold decryption mechanism. To guarantee transparency, the intermediate shuffles and decryption results, together with proofs of their correctness, are posted on the bulletin board throughout this process.
In this paper, we present a verifiable shuffle and threshold decryption scheme in which, for security parameter k, L voters, M mix servers, and N decryption servers, the proof that the end tally corresponds to the original encrypted ballots is only O(k(L + M + N)) bits long. Previous verifiable shuffle constructions had proofs of size O(kLM + kLN), which, for elections with thousands of voters, mix servers, and decryption servers, meant that verifying an election on an ordinary computer in a reasonable amount of time was out of the question.
The linchpin of each construction is a controlled-malleable proof (cm-NIZK), which allows each server, in turn, to take a current set of ciphertexts and a proof that the computation done by other servers has proceeded correctly so far. After shuffling or partially decrypting these ciphertexts, the server can also update the proof of correctness, obtaining as a result a cumulative proof that the computation is correct so far. In order to verify the end result, it is therefore sufficient to verify just the proof produced by the last server.