In India (one of the largest democracies using EVMs), security researchers are crucified if they point out flaws in EVMs, the most memorable of these events being: https://freedom-to-tinker.com/blog/jhalderm/electronic-votin...
The word machine implies predictability and being built for one purpose only. Computers are not predictable and are furthermore designed to solve any problem, including committing election fraud.
First of all, computers can be mechanical - the first computers have been [0].
Furthermore, computers are predictable. If a CPU would not be predictable, how would you program it? If you write a program, you assume that your computer is predictable and that your instructions will be carried out. (If that assumption is not met, a fault occurred - but this happens in mechanical systems as well) In fact, you can perform computation in lockstep to detect errors.
Of course, it can be practically impossible to exactly predict a network of computers with several layers of software deployed... but that is not the issue here.
The original post: http://pdt.org.br/index.php/noticias/voto-eletronico-hacker-...
I'm not even going into the 'EVMs are evil' discussion, but just pointing that this article is just a bunch of lies.
One particularly interesting case is that in one city there was the exact same number of null votes, white votes and registered voters that did not vote.
One candidate sued to have a recounting.
Not only he lost the lawsuit, but was ordered to pay a huge fine in punitive damages for "frivolous lawsuit".
Another interesting thing is that some Princeton researchers showed how to write a virus that can miscount votes on machines made by Diebold (that also make the Brazillian machines)
