I helped write the software for the Brazilian voting machines (state issued, standardized, made to spec by competing companies) and we had a long list of scenarios we had to guard against. The people who wrote the spec were field experts who studied attempted and successful (but caught) voting fraud every election. The resulting combination of hardware, software (the application itself is ridiculously simple), analysis and (and this is most important) procedures surrounding the physical devices (never left alone unguarded, clear chain of custody, created layers of protection and, in the end, a reasonably secure device. It's possible to make it absolutely secure? I'm not sure. Would that be usable? I doubt it.

It's foolish to make a flawless voting system when we can't guard against propaganda and other forms of manipulation through social media or even the most traditional paying voters (either explicitly or through promises) to vote a certain way.