A central problem in US election security is the size and complexity of ballots in most jurisdictions make automation essential in practice for accurately tallying votes. But automation (especially involving computers and software) is inherently untrustworthy (software has bugs, and can be quietly replaced with malware).
This seemed like an impossible, unresolvable tradeoff until two recent breakthroughs: the notion of "software independence" and methods for practical "risk-limiting audits"