Democratizing election verification: new methods for addressing an ancient attacker model

Abstract

Elections are a special security problem because it is not good enough for systems to be secure and results correct - they must also be verifiably so. Even leaving aside the psychological aspects (some people don’t believe evidence, or don’t understand mathematically-based evidence), many nations’ election systems fall far short of this goal. In this talk I’ll discuss a setting increasingly common in the US, Australia and elsewhere: citizens vote privately on paper, then the votes are digitized and counted electronically. Sounds simple, doesn’t it? But producing publicly verifiable evidence of a correct outcome requires carefully-designed processes. Also, running these processes meaningfully requires active involvement from the public. I’ll discuss the attacker model and process of verifiable election audits. I’ll then explain our groundbreaking techniques for auditing instant-runoff (IRV) elections and other complex social choice functions, and describe important open problems, particularly for the single transferable vote. Based on joint work with Michelle Blom, Andrew Conway, Alexander Ek, Philip B Stark, Peter J Stuckey and Damjan Vukcevic.