Reply #3: I'll give it the college try

[foo_bar]

This is fundamentally incorrect. The software is only one part of a voting process. The totality of the software, hardware and the electoral process and procedures, which include certification and testing by election officials, is what safeguards the integrity of election results. Cryptography, in particular, is only a small part of the equation.

Defensive but arguably correct. Software is a more fundamental security problem than hardware, even if human oversight is the fundamental solution. But I'd argue human trust is -the- fundamental problem.

The electoral process is designed in such a way that no single individual, or even a small group of individuals, can tamper with the election results.

Tell it to the Supreme Court. ;)

It is also important to note that such a conspiracy would not necessarily require any “security relevant flaws” in the code to accomplish its aims. Fraud of this degree would have the potential to undermine any voting system.

This is absolutely correct, but it implicitly contradicts the assertion about individuals named Katherine Harris or small groups of individuals wearing robes.

Voter verifiable receipts do have the advantages cited by the authors, though this solution essentially reduces an electronic system to a paper system, which has risks of its own. For example: Unscrupulous election officials can replace the paper ballot receipts with their own tampered copies (“ballot stuffing” in the classic sense).

Forging the original and the paper trail is more complex by an order of magnitude than simply the former. A conspiracy to do both would be unsustainably large compared to bribing a few bureaucrats.

The certification process is both rigorous and arduous. It is by no means dubious.

Famous last words.

Many states require a voting system to go through several levels of analysis before being accepted for production.

Many versions of Windows go through several hundred levels of analysis. 'nuff said.

The source code for ballot tabulation systems is generally required by statute or regulation to be placed in a third party escrow facility, to be examined only upon court order or the vendor’s failure to support the code.

They're invoking the "open source kills profit" argument, which speaks to the basic problem with making civics a proprietary trade secret.

At the polls, the system operates offline until polls close, and is then only optionally connected to upload unofficial election results to the central server.

In other words, their prototype was a star topology but they made it an "unsupported feature" once they groked the PR nightmare.

Uploading unofficial election results is done over a private point-to-point network and not through the Internet or dial-up Internet services.

Phone lines are the least of the problem. Most credit card terminals operate over "insecure phone lines" but the fraud potential lies in identity theft, not a monkey-in-the-middle modulating line noise.

The authors’ analysis provides no evidence whatsoever of the system’s incorrectness, and was undertaken without full knowledge of the election systems.

Clever use of the "sources and methods" defense. They can tell you you're wrong, but they can't tell you why without compromising trade secrets. Another good reason these voting systems should be written and audited by a consortium of Linux dorks.

This statement is based on the presumption that there is a single correct means of using cryptography. This is not accurate. The software is designed with the realization that subsequent versions will be released to address any needed improvements or requested changes; but the cryptography in the software is used as the developers intended, taking into account additional security measures, and the possibility of future development.

Marketing speak for "we're learning as we go". Cryptography isn't a magic bullet; all of the retinal scanners in the world are useless without a minimum wage security guard to make sure the eyeballs are attached to someone's socket.

To paraphrase Stalin: the black boxes aren't the problem, it's the people counting the black boxes. No amount of encryption will change a 5-4 scotus decision, so the problem is quis custodiat not the security lapses themselves.