On Thursday, September 16th I visited the Tarrant County Election Center. They were doing tests in preparation for the upcoming Constitutional Amendment election. There was a single Ballot Board Member there as an official observer and one other citizen and myself there to learn. I had the pleasure of speaking with at least four members of the staff including Heider Garcia the Elections Administrator. I wish I had recorded everyone's name, but I didn’t. Most of my discussion was with Mr Garcia and we spoke for over an hour.
I will remark that every time I’ve had contact with a Tarrant County employee, they have been very helpful and attempted to answer my questions. The County really should be commended because I see much different interactions in other counties. The staff explained the basic processes and answered every inquiry. I never detected any attempt at evasion.
They explained that the mail in ballots came in and were sorted. The ballot board opens them and reviews them for required elements - we did not discuss that process in detail. After examination the board prepare the ballots for scanning. They are scanned in one of three scanners and normally only two are in use. If a ballot is rejected for any reason a rescan can be attempted, and if still not accepted it is reviewed by the ballot board. The board is only a few feet away.
I was told that the same ballot serial number could not be scanned again once it was accepted and counted and I’m pretty certain that was demonstrated. I did not ask if it was scanned on one system and then scanned on another, if the tabulator would recognize that and reject it. I’ll circle back on that :)
When the batch is complete, the data is downloaded to a serialized thumb drive that they refer to as a “V” Drive and then the drive goes to be duplicated.
The duplicate drive goes to a workstation that is connected to the internet via a VPN to Scytle for reporting unofficial results, and the main drive goes to another workstation that is said to not be connected to the internet. I did not observe any connection to this server, but I didn’t make any special effort to do so either. This server tabulates the official results. The official and unofficial results should be identical since they are based on the same original thumb drive.
The ballot images for the mail in ballots are kept on the machine that’s attached to the scanner. They are not transferred to the thumb drive.
The data is collected periodically from the tabulation server and is taken to another workstation where it is manually entered and uploaded to the SoS site. So the data is split and sent in three different directions from three different servers. Two are internet connected and one is not. In theory anyway. We didn’t discuss it, but all three should balance.
The scanners at the precincts have a “v” drive installed in a locked and sealed compartment. There is space for two, but our current software edition only supports one. That machine arrives at a rally point, and the seal numbers are checked and the V drive is removed and duplicated. Then they make their way to the election system to go through the same process as the mail in ballots. I was very disappointed to learn that the scanners at the local voting sites do not store the ballot images. We also do not maintain the thumb drives once the data has been transferred to the tabulator.
The poll books communicate as suspected via a cellular device and actual phones have been used as backups. They communicate trough a VPN and I failed to ask where the central sever was. I presume there at the main facility. They can operate stand alone, but of course that would breed problems in an election where people can show up at any voting location. Most of the sites apparently use Verizon, but some use AT&T and the deciding factor is apparently signal strength.
I asked about the printing issues in this past election and it was explained to me to be a quality issue where the printer simply did not meet established norms. Garcia stated that they had an analysis done that demonstrated how the ink was sprayed just a bit to loose or scattered for the scanners to get a good read. I did not delve into who was or was not involved in the ballot duplication process, but I did learn a bit about it.
These ballots came from many different precincts and instead of searching out a paper ballot for each precinct, they used the touchscreen devices to generate the new ballot. Apparently keying up the right precinct on a touch screen was faster. That means that a manually written ballot was transcribed to a computer generated ballot and the computer generated ballot is what was scanned and counted. Those ballot images were not kept even though they were technically mail in ballots. I didn’t ask why. I also didn’t ask how they were paired to each other for long term identification and storage.
The local election administrators are working under the impression that the software and logs on the machines are not bound by the retention requirements. When we discussed the issue the point was made that if you were bound to keep that intact you could never enhance the equipment. I understand the argument but I’m also distrustful of the system. We did discuss the option of changing the hard drives and archiving the old ones and Garcia seemed to consider the idea. I had the idea on the way home that a bit for bit copy could be made before the drive was wiped which is mostly what we would do in an audit. There were a few times when I’d bring up a concern such as the software and he would remark why would we worry about that? We have the paper. My retort was that the public wasn’t having much success in getting our hands on the paper. I want to reiterate that none of this was a contentious discussion. It was two people having a rational conversation.
I have emailed and requested an update on the expected time frame for the software updates and I put the drive copy idea on the table.
The administrators are working with the opinion that ballot images are not required to be stored and in our case they are not stored except for mail in ballots. They have discussed capturing and storing those images in the future. The main concern that Garcia mentioned expressed, was the potential to slow scanning at a busier location. I don’t think they had tested to determine that impact yet. The local administrators also are working under the interpretation that if they do keep them, they are subject to the same rules as the paper ballots and that they can’t be accessed without a court order.
I asked about Tarrant Counties document retention policies for election materials and I was told that they were 36 months. I asked if that was written anywhere and apparently it was policy that was decided at an election board meeting. Mr. Garcia also indicated that they were expecting many FOIA requests at the 22 month mark.
I’ll make the following statements as a reasonably qualified technician with some computer and network experience, but I’m not by any means an IT professional.
I believe that this equipment is better in several ways than most of the Dominion and ES&S equipment we’ve read a great deal about. This department appeared to me to be a step above most all of the ones I’ve seen in videos, although I did and will make some suggestions for improvement.
I am not the smartest man in the room when it comes to technology, but from what I saw, a man in the middle attack, the way most all of us envision one, would be very difficult to do in the tabulation area. That doesn’t mean that isn’t possible, but it would take very inventive thinking to do it. I was almost disappointed by seeing this. Trying to balance what I have learned about how other jurisdictions worked, what we see in the data, and the topography I saw here is difficult.
I had to remind myself that as time goes on and the equipment becomes more sophisticated, that so does the adversary and we are likely looking at an adversary that has Nation State level resources. There are ways that an air gap can be jumped that don’t include WiFi. I have believed for some time that The companies such as Scytle, Clarity (recently purchased by Scytle) and Edison were the key, and today I notice that Joe Oltman is making comments to that effect as well. I don’t have the methodology figured out, but I have a few ideas.
I’m very concerned that not only here but across the country there is an attitude that the software used in these machines is not being interpreted as requiring retention. I just read the only Federal statute I could find and it’s so poorly written that varying interpretations would be easy to argue. I’m having trouble finding a full copy of the EOC 2002 version of the VSS, but the excerpts I’ve seen haven’t been iron clad either. I haven’t done enough research to know if they are binding on us here in Texas or not. I will.
I am concerned that we do not maintain the ballot images and that even if we did that they would not be published. We can make the decision to keep them at a local level, but I think that and the publication of them need to be addressed at the state level.
When something doesn’t work like I think it should, I want to take it apart and see what makes it tick.
I’m ready to do just that.