[EM] Why We Shouldn't Count Votes with Machines

[Kathy Dopp]

> "fundamental design is flawed"?  If so, obvious response is to redo the
> design.

Hi David,

The only "design" that is *not* flawed (that I know of) is
voter-marked paper ballots because it provides voter-verifiED ballots.

However the optical scanning machines that count them today are very
flawed and use no modern security, encryption, or open standard data
formats that have been available for many years. They're your basic
cheap junk, but far superior to today's basic cheap e-ballot junk.

>>
>> The flaws of DRE paper roll ballot printers include (there is a much
>> longer list):
>>
>> Studies show that fewer than 30% of voters check machine-printed paper
>> ballot roll records and fewer than 30% of voters who check (or about
>> 10% of all voters) accurately proofread their machine-printed paper
>> ballot roll records to detect any errors, so that a programmer can
>> switch up to 90% of available target votes in a way that no audit can
>> detect.
>
> I do not see how an auditor could know of and tailor the audit to the
> particular ballots the programmer did not switch.

Huh!?  I said: "so that a **programmer** can switch up to 90% of
available target votes in a way that no audit can detect."

Valid audits require manually counting ballots of randomly selected
reported unofficial vote counts.

The problem is that because only 10% of voters may accurately check
machine-printed paper ballot records, ALL MACHINE PRINTED ballots will
match erroneous electronic vote totals because 90% of the
machine-printed paper ballots can be printed to match erroneous
electronic touchscreen ballot records, and the voters would not notice
it; and no audit can detect the fraud.

Voters *could* detect the fraud, but the 10% of voters who notice that
their first ballot did not match their choices and cancels their first
ballot, may think that they made a mistake rather than the machine
when the second time they try to cast their ballot after canceling it
on the first try when they notice the erroneous paper ballot, their
paper ballot then *does* match their choices. As I said, 10% of the
ballots can *not* be switched by the programmer (only 90% of target
paper ballots and their e-ballots can be switched by the programmer),
but ALL the printed ballots will match the erroneous e-ballot totals.

This particular DRE hack was published back around May 2005 in the
Brennan Center Report "The Machinery of Democracy" and is why
virtually all (everyone I know and I have written papers with dozens
of PhD computer scientists on voting system topics) computer
scientists oppose using e-ballot voting systems with machine-printed
paper ballot records.

>
>>         Also there is a "two strikes and you are out" rule that
>> prevents the most diligent voter from having a machine-printed paper
>> ballot record that matches the voter's choices.  A voter can only
>> cancel ballot casting twice due to an incorrect printed paper roll
>> record. On the third try, the voter receives an error message on the
>> screen warning that the voter has only one more chance to cast their
>> ballot.  On the third try, the paper roll ballot record whizzes
>> quickly inside the canister WITHOUT GIVING THE VOTER A CHANCE TO SEE
>> THE PAPER RECORD!
>>
> What does it matter?  How come the redesign failed to attend to properly
> recording the vote?

I do not get your question. If you want to know more about this
particular "Two Strikes You're Out" flaw of DRE-printed paper ballot
records, either:

1. If you personally vote on a DRE, try cancelling your ballot twice
and then see what happens when you cast your ballot on the third try.
(Take a picture of the warning screen with your cell phone before
pushing the button, and then watch the ballot quickly roll up before
you can see what is on it.)

or

2. Read the NJ Institute of Technology studies of DRE printers which
caused NJ to refuse to certify any of the DRE paper printers.

Without a limit on the number of times a voter can try to print a
matching paper ballot record, and without a way for the voter to bail
out of casting a vote on a DRE which refuses to create an accurate
paper ballot record, then obviously there would be other problems,
like running out of paper in the paper rolls (poll workers frequently
have problems loading the paper, load it backwards so it does not
print, and the papers frequently jam while printing, or keep the
covers closed so voters don't see the paper ballot records, or voters
can easily sabotage the paper so that it appears to work during the
elections but all the records are erased at the end of the election.
(See the CA SoS study of voting systems.)

Sigh, so the FLAW is the inanity and expense and hassle of trying to
keep a printer running in every polling booth during elections rather
than using a less costly paper ballot, and more importantly the fact
that the machine, rather than the voter is marking the paper ballot
record.

Fix that cannot be done without switching to voter marked optical scan
paper ballots.

>
> Huh?  There seems to be general agreement that present DREs need replacing.
>  I only ask that we try for usable replacements.

Over 60% of US election jurisdictions already implemented "usable
replacements" - the optical scan paper ballot system.

> Crane's paper does not explicitly mention Shamos or its authors having such
> a degree - it does mention involvement in voting.

Not surprising.

ANY SYSTEM can be assumed to be inaccurate if it lacks a routine
method for detecting and correcting errors that is independent of the
software. Since voting is much more difficult to secure than banking
due to the secret ballot, the only method of providing software
independence is a paper ballot.

You certainly would not expect me to deposit my vote anonymously
without a receipt into a bank, so why would anyone want me to use such
an insane method for voting which would make it trivially easy for any
insider to alter the election outcomes undetectably?

>
> For one Crane author, Edward Cherlin, I read of working on affordable
> software and hardware for voting around the world.
>     Hopefully he is intending such to be adequate.

Many good folks are working on affordable, open source optical scan
voting equipment because most of today's optical scanners are also
hackable junk.

As Ion Sancho of Florida said (paraphrased), "It will take a long time
to fix our voting systems, in the meantime we need to push for
AUDITS."

>
> Do not know if such a degree was available when I was in college -

How long ago were you in college? Computer science degrees have been
around since atleast 1969 and I believe long before then.  That's when
I took my first computer science course. I have a couple of friends
who learned to program when they had to hard wire computers to program
them.


>  could not
> have learned much presently usable.

I first programmed with punch-cards.

> Remember a tidbit about weather
> forecasting.  Could give a computer data to predict for tomorrow - by the
> time program was done you could look out the window and see if it got it
> right.

:-) Well that used to be true, but weather prediction programs are
much more sophisticated and faster nowadays.

>     Remember a problem later at work - too much data for Fortran to fit in
> available memory.

It is easy to overrun computer memory  with data and programming
instructions, depending on the application and the skill (or lack of
it) of the programmer. I would imagine that the new weather programs
take a huge amount of memory to run.

> Heard of a new language - Jovial.  Available staff was
> engineers who could hardly spell "program" and assembly programmers who
> could hardly spell "compiler".  I successfully installed compiler and taught
> staff for project to use.
>     Later - task could not execute in available time - invented new, faster,
> instructions then used in available computers.

Yes, well the Internet was invented so that various researchers could
use others' faster computers at different locations to run
long-running programs.

Cheers,

Kathy