I'm sorry if this comes off as a bit hostile, but did you read at least the title of the article? "DARPA is building", not researching.

The content of the article reflected this - mostly about actually making a system, very little about research on the math behind secure voting on untrusted hardware. It's clear they want to use this.

I'm sorry, do you know what the acronym D.A.R.P.A. stands for, or that they explicitly stated they aren't the production implementer of any such system? Do you know what the network you're throwing packets at is based on?

Anything that relies on secure hardware/software is broken, since those are impossible to assure. The only part that can actually guarantee secure voting is everything that's outside the hardware. Things like cryptographic signatures, hashes, paper trails, whatever. Things that don't rely on the black box you're presented with when you vote to work how you think it works.

Researching that doesn't require a prototype. So why are they building one? As a first step to using a voting system based on it. This isn't 'research', it's propaganda to sell the idea that hardware can be secure. They'll give it to Defcon or whoever, and once they fix all the weaknesses hackers can find, they'll triumphantly declare it's 'secure', and that we can switch to e-voting, and not to worry our pretty little heads about chip/compiler-level backdoors, or if the builders of the system themselves subvert it.

They can state they won't be the implementer of such systems all they like, that doesn't change the intent of this program - to push e-voting reliant on 'secure' hardware.

DARPA isn't a voting system end user, and things built by DARPA tend to be proof-of-concept things that explore the possibility space and help later users set requirements for follow-on procurement.