Home > Politics, Tech, Voting > Report on my California Primary voting experience today.

Report on my California Primary voting experience today.

Nothing horrible (this time!), just a description of my lovely travel to/from the polls, followed by a bit of detail, which includes minor criticisms. Read on for more …

Flash! There actually IS something horrible I discovered as I (almost) finished writing this. It’s way down at the bottom. If you have election machine and/or electoral integrity expertise, if you do nothing else, scroll down to the bottom! Now, back to the regularly scheduled blog post:

I took particular joy in travelling to and from the polls by bike. The weather was clear, a tad cool, but I was still comfortable in shorts. The wind was also a tad higher than I’d prefer, but not so much as I would have felt I was fighting the wind. I always take joy in exercising my right fulfilling my civic duty to vote. The joy of a pleasurable bike ride just added to it. I even tweeted about it:

And yes, this is the blog post I promised in the tweet!

Now onto the actual voting …

The precinct was nearly empty when I arrived – only the 3 precinct workers and one other voter. They found me quickly in the poll book, I signed, and they asked me if I wanted to vote on the touch screen ballot marking device (BMD) or hand mark a paper ballot.

This is the first thing that troubled me. I’d previously investigated the new machines my county is using for the first time today, so I knew that the “ballot” produced by the BMD has serious flaws. It’s not the same in form or layout as the paper ballots. Instead, it’s a single piece of paper listing how you voted (but omitting all the choices you didn’t make), with a big 2d bar/code that, supposedly, is a machine readable encoding of what is also printed for you to read and verify that the BMD correctly created your ballot.

Folks, this is a problem on a number of levels.

First, as most of you know, I’m a firm believer in hand marked paper ballots since they are unhackable when marked in precinct and placed by the voter into the ballot box. I acknowledge that BMDs of some sort are necessary to accommodate voters who, for whatever reason, are either unable to hand mark a ballot or face significant difficulties hand marking a ballot. But BMDs should only be used to fill that specific need, not offered to every voter!

Next, based on a conversation I had about 6 weeks ago with the person who programs the voting machines (both BMDs and Optical Scan Devices) used in our county, I know there is no way to prove that the votes encoded in the bar code actually match what is printed on the page! There exists a “verification scanner” that can scan the barcode and (purportedly) tell you how your encoded vote will be tallied, but there are a couple of problems:

  1. These verification scanners aren’t in every precinct! They’re expensive, so the county only keeps a few centrally for use in hand recounts. Thus they’re useless as a means for a voter to verify the barcode before surrendering their filled ballot to the combination Optical Scan Device/Sealed ballot box!
  2. Even if they were available in every precinct, they still don’t guarantee that your ballot will be tallied identically by the OSD and the verification scanner. The two devices run on different hardware and are controlled by different software. And both the hardware and the software represent hacking vulnerabilities. Further, the county programmer I corresponded with admitted that there’s no way to prove to the public, in real time, that the ballot they just checked with the verification scanner will be tallied by the OSD exactly as the verification device reports.

So voters easily capable of hand marking a paper ballot are being offered a shiny, high tech alternative BMD which is part of a system that makes even the paper trail they create unverifiable – even when the voter thinks they’ve verified it!

Sigh. On to other things.

My precinct had one – yes only one – voting “station” that can accommodate at most 4 voters at any one time. Here’s a quick and dirty model I made in SketchUp – as I didn’t think to ask if I could take a picture:


Weird – in the post editor I’m using to compose this post, the model actually appears above, but when I tried to “preview” it, only the link is there. So if you see no model and are interested enough, please click the link!

The writing surface on these “stations” is too small!! The ballots are 3 11″ x 17″ pages with choices to be made on both sides of all 3 pages. As you’re marking one page, what do you do with the other two pages? And what if you’ve brought a filled out sample ballot – now you have 6 large pages you’re juggling while trying to mark your actual ballot. I fell somewhere in between – I made myself a cheat sheet, in ballot order, listing my intended choice in each race/ballot question. And I abbreviated the offices too. I ended up with a cheat sheet that, when folded into quarters, had ballot races on one side and propositions on the other, yet I still felt a tad overwhelmed literally shuffling paper! I could easily imagine someone simply overlooking one (or more) of the 6 sides of the ballot we had to complete.

I even made a mistake that required me to void one of my ballot pages and get a new one from the precinct workers! I had two races where I voted for a write-in candidate. On the first one, I accidently used the write-in space for the previous race – on which I’d also chosen my preference! I’m not sure how much the smallness of the writing surface contributed to my making this mistake, but it wasn’t zero. Had I not noticed it, my choice for the prior race might have been rejected as an overvote and my choice for the race I wanted to use the write in counted as “no vote” i.e., an undervote! (more on this later).

Now, with my ballot filled out completely and (after double checking) reflecting my exact preferences, with no undervotes, overvotes, or accidental other markings, it’s time to feed it into the OSD. The OSD allows feeding in the ballot pages in any orientation (that’s good) and has a screen that offers a choice of languages (which I didn’t explore). Once you tap the on screen “Start” button, you’re prompted to enter a ballot page then, while it’s “processing” it rejects any attempt to enter the next page. Per the precinct workers, it will also reject ballots with overvotes, accidental markings, and markings near the ballot bubbles that they can’t decipher. It can’t detect an undervote as many voters choose not to vote in some races. I didn’t attempt to confirm any of this.

After this, I grabbed by “I voted!” sticker and left the precinct. That’s when I asked an arriving voter to do me the favor of taking the picture I included in the tweet.

Thoughts that formed after I left the polls:

  • Although my precinct isn’t normally busy, there are times – especially during evening commute hours – when you can reasonably expect 10-15 voters to be there waiting to vote. The combination of only one 4 position voting station with the routine offer to use the BMD leads me to bet that there will be a few voters who choose the BMD solely to avoid the line for an available position at the voting station. We should never make it more convenient to vote using a BMD than it is to hand mark a paper ballot.
  • I totally forgot to ask if the OSD saves the scanned ballot images! Much less if it saves only ballots with write-ins vs all ballots. Much less if the ballot images would be made publicly available through a website within a few days – a transparency measure I consider essential to trusting OSDs to accurately decode the ballots and correctly tally the votes.

I have more I intend to say on a few issues:

  • Why I say I consider that public posting of scanned ballot images is critical to transparency when humans don’t first hand count the ballots.
  • A possible – I emphasize possible – idea I had that might resolve some of the non-verifiability issues of the 2d barcodes produced by the BMDs.
  • The beginnings of an idea of how blockchains might reasonably be used to help ensure electoral verifiability and transparency.

But I’ve too many other things I want to get done soon to include them here. And given the rarity of my blog posts these last few years, I’m not going to promise to actually expand on them. If you really want my thoughts on these, wait a couple of weeks to see if I manage to do a post on them. And prompt me by comment here if I haven’t!

With all that said, it’s time to invite you to comment below!

And “Ta Ta for Now” — Steve

Oops! I almost forgot! For you voting integrity geeks who read this, our new OSDs are “Imagecast Evolution” machines made by Dominion Voting (Does that company name make you shudder? It sure makes me shudder!)

I couldn’t find the actually identity of our county’s new BMDs, but digging into the county’s website yielded a couple a video. If somebody who might recognize the machine watches this and does know the details, please comment!

A final “Oh Aitch Eee Double Hockey Sticks” moment:

I just now, after writing all of the above, noticed a detail on the Dominion Voting website. Believe it or not, our OSDs can also double as BMDs! About half way down the page, buried in a bullet point you’ll see:

  • Can mark and scan ballots up to 22 inches

The color of the word “mark” isn’t in red on the page, I changed it for emphasis.

Wait. What? Oy Vey! Another detail that might be a vulnerability that at least needs investigating: Could a malicious entity cause the programming to add a mark to another candidate in a race where the programmers don’t want votes counted for the candidate I voted for? Could this occur AFTER the check for overvotes? Could the machine tally it as an overvote while simultaneously making my original ballot show an overvote that I didn’t put there? That seems to me to be an awfully big vulnerability. It changes the original AND the tally. It wouldn’t be discovered in a hand recount! The only way to possibly discover it is to read, decompile, and analyze the actual code used on voting day on every machine!

Now I’m going to have to take time to make sure some voting integrity experts know what I’ve apparently found. So let’s try this again:

Ta Ta for Now – Again!

PS: Apologies for any typos or unclear thoughts. I’ve spent far longer than I intended composing this, and it’s far longer than I expected when I started. I simply don’t have time to do a good proofread right now.

  1. 05 Jun 2018 at 18:59

    Thanks for the detailed item, Steve. No time to dig into it for the moment to confirm (please feel free to do so with your Registrar!), but I believe the “ImageCast Evolution” is a series of machines meant to work together. So, I can’t say whether I know or not if the scanners can also print (as you suggest) or if they are separate pieces of machinery. Worth asking the Registrar about, however, as I’d love to know! (If THEY even know, that is.)

  1. No trackbacks yet.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: