Philip Stark
UC Berkeley statistician who introduced risk-limiting audits in 2008 — a way to check an electronic election result by hand-counting as few paper ballots as possible.
The idea is easy to state and hard to do the maths for: draw ballots at random from the paper record and compare them with what the machines reported, until it is statistically unlikely enough that the wrong winner was declared. A wide margin needs a few hundred ballots. A narrow one needs more, up to a full recount in the worst case. What you get is a guaranteed upper bound on the chance that a wrong outcome survives — whether the cause was a bug or an attacker.
The point is that you don't have to trust the software at all, only the paper and the dice. Colorado became the first state to run them statewide in 2017. Security people had been saying "keep the paper" for years; Stark answered the obvious follow-up of what to actually do with it.