What proof do we prefer? Variants of verifiability in voting

In this paper, we discuss one particular feature of Internet voting, verifiability, against the background of scientific literature and experiments in the Netherlands. In order to conceptually clarify what verifiability is about, we distinguish classical verifiability from constructive veriability in both individual and universal verification. In classical individual verifiability, a proof that a vote has been counted can be given without revealing the vote. In constructive individual verifiability, a proof is only accepted if the witness (i.e. the vote) can be reconstructed. Analogous concepts are de- fined for universal veriability of the tally. The RIES system used in the Netherlands establishes constructive individual verifiability and constructive universal verifiability, whereas many advanced cryptographic systems described in the scientific literature establish classical individual verifiability and classical universal verifiability. If systems with a particular kind of verifiability continue to be used successfully in practice, this may influence the way in which people are involved in elections, and their image of democracy. Thus, the choice for a particular kind of verifiability in an experiment may have political consequences. We recommend making a well-informed democratic choice for the way in which both individual and universal verifiability should be realised in Internet voting, in order to avoid these unconscious political side-effects of the technology used. The safest choice in this respect, which maintains most properties of current elections, is classical individual verifiability combined with constructive universal verifiability. We would like to encourage discussion about the feasibility of this direction in scientific research.