[ale] Its over. Maybe

[aaron]

On Thursday 04 November 2004 05:14, Michael Still wrote:
> Ok... I need to clear some things up..
> 
... and a clarification / correction  of your clarifications, from research 
and confirmation as a Poll Watcher at the Gwinnett county tabulation center 
Tuesday night:

> After the election is over a worker then pulls a pcmcia style flash
> drive out of all the machines and drives them to the county election
> office where each card is read into a server that counts up the votes.

Each precinct does drive the DRE PCMCIA cards to the tabulation center, but 
FIRST they read all the cards for the precinct into one DRE machine that's 
been designated as the "accumulator" system and is equipped with a MODEM. The 
votes cast counts are noted with the other precinct materials, and the 
election totals are sent via the Modem in to the polling station central 
system for "instant" reporting. Once the cards are delivered to the county 
center, the precinct materials are checked in, the paperwork reviewed, and 
the serialized PCMCIA cards accounted for. 

> If his car crashes and burns up then the votes can still be read from
> the diebold machines from one of at least two other long term storage
> devices.

This is not correct.  Besides the PCMCIA card, there is no other complete, 
independent, long term DRE vote data processed and retained by the voting 
machines. The DRE machines have a single ballot memory space, but that is 
cleared at the beginning of each voter session, so it is obviously not "long 
term".  The systems do store an internal flash memory record of all the voter 
actions on the touch screen,  but this is basically a compact list of the X/Y 
coordinates touched by the voters. While the data can be used to re-create 
the ballots and votes, it is not a complete or independent record. The 
Secretary of State claims this record is the part that meets the "paper 
record for manual audit" requirements of HAVA law, but none of the standard 
precinct and accumulation procedures access this data or attempt to recreate 
printable ballots. 

FYI,
My notes from Gwinnett show that, of 148 precincts, there were more than a 
dozen cases where precinct poll workers had difficulty accumulating and / or 
phoning in the data for all their cards, so when their cards were delivered 
they were read into the Central System on a precinct by precinct basis via an 
adjacent, Modem connected accumulator DRE.

Over the course of the night I observed the workers having problems getting 
several of the memory cards to read and they even had some would crash the 
DRE system being used to feed the data into the Central computer. Also, some 
cards that had been read in to the Central database later showed up as unread 
and had to be re-sent. (-; Of course, whether the Diebold software also 
totalled those votes twice is anyone's guess. ;-)

Vote totals from the County Central system are periodically sent by Modem to 
the State's central system, which is on Network (or at least it was for the 
2002 elections).

The person operating the central system has access to enter vote counts 
manually, and actually does this for the Optical Scanner (paper tape) totals 
of Absentee Ballots and totals of Authenticated Provisional ballots (which 
are treated as 2 additional precincts). All voting records are stored in a 
common M$ Access database file and totals can be edited by anyone
reasonably familiar with the M$ Access software. 

While there are many cross checks to confirm that the numbers of ballots cast 
and submitted match up, there is nothing in the system that allows monitoring 
or auditing of the actual vote totals.

> The pcmcia cards are marked with a hash before the election and then
> checked after the election to ensure the card is authenticated.
>  The  server is not connected to a network, runs win2000, and does nothing
> all year until election day.  The county election officials are
> instructed to lock the machine in a secure room, but I'm not sure if
> this is enforced. 

In Gwinnett the Central Tabulation system was in a small, separate room with 
an observation window. It is connected to 2 DRE accumulator systems and a a 
Modem bank of common carrier phone lines... I think they said 23 lines were 
available. The I.T. tech said they had challenges with the lines during 
testing because of heavy phone traffic into the registration office the week 
leading up to election day, so the lines were obviously on the common 
building trunk.

> I would be sure to guess that in the bigger
> counties it is locked up. The server has a tripwire type system check
> that it goes through as well.  I'm not sure where the data goes from
> there, but I am pretty sure it does not go to a central system via a
> network of some sort.  My guess would be another pcmcia style card
> that is mailed back and forth using the same hash style authentication
> all the way to Cathy Cox's own system.

(see above)

> Cracking the Ga voting system using technology (or subverting) is not
> as easy as you would think.  The system may still be vulnerable, but
> it would take a major effort.

...but the fraud could be effectively carried out by as few as 1 or 2 people, 
is especially easy for Diebold insiders, could easily be done in a way that 
was undetectable, and could effect the election outcomes of an entire State 
or Nation.

peace
aaron