Vote Verification using Hard AI Problems

Recently proposed end-to-end independently-verifiable (E2E) voting schemes provide encrypted paper receipts to voters, who may later check that these receipts are in the electronic ballot box. Unfortunately, few voters are likely to follow up on the voting process after leaving the voting site; as a result, few receipts will be checked. This paper describes an enhancement to E2E schemes that does not require the voter to perform a task outside the polling booth. It enables the voter to electronically transmit her receipt, from the polling booth, to a trusted external verifier. This is done through the use of a human-verifiable digital signature primitive whose (short-lived) security depends on the hardness of an AI problem. The primitive enables the voter to be certain—without access to trusted computational power in the voting booth—that the receipt has been securely deposited with the external verifier. The approach presents several advantages: the voter is not required to do anything outside the polling booth, no receipts are needed after polling, all receipts generated by the polling machine can be checked, and any classical digital signatures on receipts can be checked instantaneously by the trusted verifier. Additionally, an audio-based format is an easy extension for those with visual disabilities. The cost of these benefits is the introduction of the verifier, who needs to be trusted not to launch a denial-of-service attack.