"Software independence", posed by Ron Rivest in 2007, formalizes a design goal for a voting system: A software error (even malicious software) anywhere in the system must not be able to cause an undetectable, unrecoverable change to the correct outcome of an election.
Risk-limiting audits, invented a few years later by Philip Stark, provide a practical way to achieve software independence with a type of existing voting technology (optical-scan paper ballots).