Software Support for Risk-Limiting, Post-Election Auditing

Abstract

The security of our government elections has become a topic of enormous concern, especially in the wake of the 2016 presidential election. Serious security and privacy flaws in some of our most widely used voting machines can compromise the integrity of our voting system. The most effective way of verifying the reported outcome of our elections is by manually auditing physical copies of the ballots. Since auditing every ballot is unrealistic, we use a number of audit algorithms to determine a subset of ballots to audit. Current auditing laws, however, still often require auditing many ballots, which is both expensive and inefficient. Risk-limiting, post-election audits can improve efficiency in election auditing by examining fewer ballots, while still providing statistical assurance that the reported election outcome is correct. We aim to encourage the use of risk-limiting audits by creating software that will guide government officials smoothly through the audits. Such an application would both minimize human error, and create a unified and verifiable process of auditing election results.