Adapting Helios for provable ballot secrecy
Résumé
Recent results show that the current implementation of Helios, a practical e-voting protocol, does not ensure independence of the cast votes, and demonstrate the impact of this lack of independence on vote privacy. Some simple xes seem to be available and security of the revised scheme has been studied with respect to symbolic models. In this paper we study the security of Helios using computational models. Our rst contribution is a model for the property known as ballot privacy that generalizes and extends several existing ones. Using this model, we investigate an abstract voting scheme (of which the revised Helios is an instantiation) built from an arbitrary encryption scheme with certain functional properties. We prove, generically, that whenever this encryption scheme falls in the class of voting-friendly schemes that we de ne, the resulting voting scheme provably satis es ballot privacy. We explain how our general result yields cryptographic security guarantees for the revised version of Helios (albeit from non-standard assumptions). Furthermore, we show (by giving two distinct constructions) that it is possible to construct voting-friendly encryption, and therefore voting schemes, using only standard cryptographic tools. We detail an instantiation based on ElGamal encryption and Fiat-Shamir non-interactive zero-knowledge proofs that closely resembles Helios and which provably satis es ballot privacy.