Abstract
In this paper, we analyse information leakage in Ryan’s Prˆet `a Voter with Paillier encryption scheme (PAVPaillier). Our analysis shows that although PAV-Paillier seems to achieve a high level of voter privacy at first glance, it might still leak voter’s choice information in some circumstances. Some threats are trivial and have appeared in the literature, but others are more complicated because colluding adversaries may apply combined attacks. Several strategies have been suggested to mitigate these threats, but we have not resolved all the threats. We leave those unsolved threats as open questions. In order to describe our analysis in a logical manner, we will introduce an information leakage model to aid our analysis. We suggest that this model can be applied to analyse information leakage in other complex mixnet based e-voting schemes as well. Furthermore, we introduce a simplification of PAVPaillier. In our proposal, without degrading security properties such as voter privacy, verifiability and reliability, we no longer need to apply the homomorphic property to absorb the voter’s choice index into the onion, thus we step back to employ the ElGamal encryption. This results in a simpler and more straightforward threshold cryptosystem. Some other attractive properties of our proposal scheme are: unlike traditional Prˆet `a Voter schemes, the candidate list in our scheme can be in alphabetical order. Our scheme not only handles approval elections, but also it handles ranked elections (e.g. Single Transferable Voting). Furthermore, our scheme mitigates the randomisation attack.