The Helios mock election in early 2010 was generally successful with 379 IACR members casting ballots. Several issues were identified as necessary or desirable to be improved before use in an actual IACR election.
The primary issue was the requirement for Java on the Helios client. Several members were unwilling to install Java on their own machines or to use alternate machines on which Java had been installed. To alleviate this concern, a feature was added to Helios which enables ballot encryption to be performed on a Helios server. There is a trade-off here because use of this feature gives the Helios server access to a raw ballot which could, in theory, compromise privacy. However, this amelioration was deemed adequate by the committee.
A desirable improvement which has not yet been implemented is full threshold encryption. The current implementation of Helios allows for generation of a key shared amongst trustees, but all trustee shareholders must cooperate to perform the final decryption of the tally. This creates a robustness concern since if any trustee loses a key, the election cannot be completed. The long-term solution to this problem is an implementation of true threshold encryption wherein not all trustees will be required for final decryption. However, the committee has deemed an interim solution of careful key backup by each trustee to be adequate for the short term.
Another desirable improvement is to have multiple administrators for an election. The election administrators can share a single credential for administration, but it will ultimately be desirable for election committee members to be able to each use their own credentials. This has been deemed to not be a short-term concern.
One final improvement which we would like to see is an ability to randomize the order of candidates on each ballot instance presented to a voter. This would essentially eliminate bias due to a fixed candidate ordering. While we would not use this feature in 2010, it would be nice to have the additional ability to collect statistics on the ballot placement of candidates selected by voters to further research on the effects of ballot ordering.
Overall, Helios v3 now meets the needs of the IACR, and we see no obstacles to using it for IACR elections. We look forward to future improvements which will enhance the Helios experience in the future.
See more details here.