Physical collection is risky, but we can mitigate the risks.

I can't speak to the US system, but here in New Zealand every political party is entitled to appoint scrutineers to each polling place. They cannot speak or interact with anyone but they watch the whole process from checking voter entitlement, through transporting the ballots and then counting.

The idea is each party distrusts the others so won't let them get away with rigging the ballot. Mutual distrust produces a trustworthy outcome.

Election administration is by state in the US. To my knowledge, every US state except for West Virginia allows some form of "observer" or "challenger" (or sometimes both with different purposes), often appointed by either political parties or candidates. It's a bit of a patchwork from state to state, but yes, in general, there is a system of interested persons observing the polling place.

(In this state, observers and challengers are different, with challengers specifically serving the purpose of challenging individuals that may not actually be qualified to vote. Since the pollbooks are electronic today challengers are rarely seen, they were generally only able to challenge clerical errors that are no longer seen with computer pollbooks)

Kind of hard to observe electronic voting.

If someone wants to screw with physical votes, they have to access the boxes which hold them. Easily detectable.

Electronic ... you can have the software changed to modify votes and few would be the wiser - even have the changes hide themselves if you are smart.

Electronic votes can be checked statistically after-the-fact through audits, but that's about it.

It's mind-boggling we don't do that already -- most voting districts would need only a few random ballots, and you can gradually increase your sample size and check again before needing to trigger a full recount.

But then you would have to trust the software doing the checking.

This seems overly simplistic. Both have ridiculously tough to harden attack vectors.

In physical, it isn't enough to protect where the votes are done. They have to be transported back and ultimately counted somewhere. At any point in that process, they are vulnerable. No?

>In physical, it isn't enough to protect where the votes are done. They have to be transported back and ultimately counted somewhere.

No, they don't. At least in my country they are counted in place.

All the observers sit next to the ballot for the whole process, and when the voting ends, the box is opened and the votes are counted. The observers being several people appointed at random (like a jury) to check vote credentials and count the votes at the end + appointed representatives from each party present.

Everything is recorded on paper, the votes are stamped and kept, and the tally is then reported for that voting district.

Even if the people appointed at random wanted to tamper, they'd have to work all together + get the party representatives to agree with it, because it all happens in the open, ("reading vote #N, says party X, do we agree it says party X? (shows the vote around) registering vote #N for party X (people look as the vote is recorded, two people sign next to the vote's registration)).

Usually after the ballots are closed (election ends), it takes 5-10 hours for those people to count all the votes for a district. Then the number is announced. All country districts are announced publicly, so any individual party representative or "jury" member of any district can challenge if the numbers announced (and used for the final country-wide tally) are not accurate.

They are not allowed to leave the room, and there's also a policeman present outside.

I've been an "Election Judge" twice in Chicago. I administered my polling location twice. After voting ends, we physically collect the votes and transport them by our own vehicles to a polling location to be counted.

Well, not in my country. We do it in place.

Which proves that "votes have to be transported to be counted" is not some inevitable byproduct of the paper-voting process as the parent made it sound like.

Perhaps Chicago could adopt counting in place?

So, how many agents are we talking about here? How are they counted at the location you too them to? Two easy targets to identify.

I'm not claiming they are impossible to harden. Just not as easy as people are claiming. And super expensive. Such that if you were truly intent to defrauds place, you would focus on poor sections first.

And our best method of defense is probably our extensive polling tests nowadays. The more we have, the more corroborating evidence we have to an outcome. This protects both forms of counting.

Vulnerable to what, though? Physically changing votes takes time, and swapping them leaves a paper trail. If someone miscounts paper votes, you can just recount them.

With a computer, you flip a bit and there's no record. Votes are miscounted? Tough, those numbers are a real as any other numbers. And how much time does it take to swap a vote? Less than a microsecond?

> With a computer, you flip a bit and there's no record

That’s a turn of phrase I didn’t expect on HN.

Do we work in a field where programs don’t have logs, gateways don’t exist, checksums and securing data integrity is not a thing ?

I think HN understands better than most that any digital data is fragile and ephemeral. Theres entire fields around just preventing tampering with it.

But then do we understand as well how physical data works, what's the actual shortcomings etc. ?

Personaly I don't think I do, yet even at my personal level I have anecdotes of ink just fading out of paper, or countless of widespread voting frauds from decades ago.

I have the feeling we are putting paper and physical media handling to a higher standard because we don't know as much about it.

Yes. Physical data is well understood. Inks fade, so you use a different formula and keep it out of the light as much as possible. Inks use chemicals, so even if it’s not visible you can still see where the writing was done, inks are pressed into the paper and change the physical structure of the paper in the process.

It takes a concerted effort to change paper ballots.

The issue is not physical data though. We are talking about a voting system, with agents, suppliers, observers, ballots and people handling them.

For instance some paper elections in Africa have crazy high voter prticipation when not so many people showed up.

That’s an extreme and we could point the finger at blatant corruption. We’re not at these extremes, but where are we on the spectrum?

For instance we don’t have any clear idea of how much corruption we have, to the point that “perceived corruption” is the best approximation.

What I’m going at is, to evaluate how much trust we put in an electronic voting system, we’d need better views at the current system than “paper is better because it’s physical” (that’s not your argument, I take a less nuanced position as example)

> a field where programs don’t have logs, gateways don’t exist, checksums and securing data integrity is not a thing

Yes, that's the field of voting machine construction - it's lowest-bidder garbage built by and sold to people with no concept of reliability.

It's so bad that Diebold had to spin off and rebrand their voting machine division, out of fear that security and reliability issues with the EVMs would tarnish their main ATM business.

The ATM security is also pretty bad (and constantly plagued with skimmers), it's just that you can refund money but you can't refund votes.

We also work in the field with a history of all these measures being circumvented.

The closest parallel we have is DRM, and the track record there is.. less than stellar.

Picoseconds or at most a couple of nanoseconds

Same in Italy, each party can have one person observing proceedings. There used to be tons of parties, hence there were lots of eyes around. The better-organised parties also tracked in realtime whether their sympathizers had actually voted, seat by seat, and could send messages through relays to hurry them to the booth. Before computers, journalists typically relied on the major parties to have the first results, well before they were officially declared.

This has changed in modern times, as a number of parties disappeared after electoral reforms; and the risk of tampering and shenanigans has increased substantially.

Also on this point, we know the large silent majority is made up mostly of moderates who want to live in a civilised democratic country where the government either leaves everyone alone or have a mass benevolent impact.

As long as attacks don't scale it seems safe to assume corruption will be localised and the integrity of the system will hold. The risk is that IT is used to centralise democracy to bring down costs, then becomes compromised in an unrelated attack.

But if the parties are picking partisan voting officials to oversee the election, there might not be much moderation or civility among them.

Yeah nah, that's choice but, at least our scrutineers aren't so easily bribed by fish.

Edit: should read articles before commenting..

It's animal cruelty to open seals with shims, no matter what they're protecting.

So what happens if one of them sees something fishy? Can they report it in a way that will lead to action?

They can collude to exclude a third party.

This is essentially how the Directory managed to maintain power during the French Revolution when their party was set to lose a popular election in 1798. In departments where the Directory was expected to do especially poorly, they engineered electoral irregularities ahead of the vote that could then be used to invalidate the results. This became known as the Coup of 22 Floréal.

> Here’s a unique idea: if you want to influence elections, tamper with tamper-evident stickers to get votes from an unfriendly district thrown out.

Because you don't automatically throw out votes even if the seal is tampered.

You have other means of cross-verifying authenticity. You can look at the voter roll signatures to see if the vote totals match. You can do statistical analysis versus the expectation and look for anomalies. You can ask observers if they saw anything untoward. etc.

Paper is secure NOT because it is untamperable. Paper is secure because we can bury it a whole host of interlocking cross-checks--of which tamper seals is one of.

Paper is also more-secure because tampering costs more in terms of time/money/energy/risk.

It's much easier to remotely and anonymously mass-rewrite electronic records than to physically move the atoms of paper and ink.

And paper is nice cause lay people can easily understand and verify the system.

How do you prove that an electric system wasn't tampered with? How do you verify the voting machines actually run the verified code? How do you verify that some sub-component didn't hack the RAM?

Winning an election is too valuable and the risks to myriad for me to ever trust electronic voting. Especially as as you mentioned tampering with paper on a large scale will likely leak something went wrong, whereas electronic tampering might never be discovered and can be accomplished at large scale.

To me, the problem with electronic voting seems to be that in order to trust the system, it is hard to avoid giving up anonymity.

Consider for example using blockchain for this. Every eligible citizen gets a "VoteCoin" from the voting officials and deposits it in the official "VoteWallet" of their party of choice. At some cutoff point no more transactions are allowed or considered.

Each voter could verify that their vote went to the right party, and the voting officials could easily verify the votes (no unknown VoteCoins for example).

However now everyone knows which party else everyone voted for...

In general the issue seems to be that if someone voting uses an electronic system they cannot rely on what that system reports back to them. It could be hacked to show whatever. And in order to remain anonymous the person voting has to be the one that verifies that their vote went to the right party.

There's been a lot of interesting research on ways to do this securely. Some of it is listed here on Wikipedia [1].

1. https://en.wikipedia.org/wiki/End-to-end_auditable_voting_sy...

> It's much easier to remotely and anonymously mass-rewrite electronic records than to physically move the atoms of paper and ink.

Exactly.

Hacking paper doesn't scale.

Is that really what happens? I would hope that if enough votes to come within a margin of error of changing the outcome were thrown out, the ballots would be re-cast. Or is there enough anonymity to make that impossible?

This is absolutely doable, and is really the logic behind most current physical election security procedures. Procedures are designed less to prevent tampering than to make tampering obvious. Re-voting can theoretically be done, because (public, real-time) records are kept of who voted at a particular polling place, not counting provisional ballots - look outside for a "street index" at your polling place.

(Source: am volunteering for the second time working the physical polls.)

I'd guess that repeating a vote favours whichever party has a particular demographic (younger?, less employed??) or whichever is better funded (to encourage/assist getting to polls) that subsequent votes get less and less attendees?

So, forcing a revote might be a way to swing a marginal seat in your favour.

As far as I know, elections can't be redone on a federal level. I could be wrong, but I certainly don't know of any situations where they have been.

They absolutely can be: see https://www.archives.gov/federal-register/electoral-college/..., specifically the section of the US code that speaks to "Failure to make choice on prescribed day" and "Determination of controversy as to appointment of electors". Basically - if for any reason a state can't pick electors on election day, it can do so afterwards by whatever (constitutional) method it desires; and if there's a disputed election and a state has a (pre-election) law enacted for deciding controversies that law just has to come to a decision within 6 days before the electors meet.

The physical collection box is generally (always?) out in the open so tampering would be obvious. Even with electronic voting machines it is rare to get enough privacy to hide attempting to tamper with the back of the machine like this article discusses. Voting machines without a user verifiable paper trail are certainly a problem, but they pose a much lower risk to the health of our democracy than other issues like what is currently happening in Georgia.

The major concern surrounding voting machine or ballot tampering is usually at early voting locations, since there it is typical to leave the ballot boxes and/or voting machines unattended overnight. The state I'm a poll worker in does have reasonable additional precautionary measures, such as printing and retaining daily open and close reports from the digital tabulators to guard against overnight ballot stuffing, but this does present an opportunity to do something like tampering with the machine so that it incorrectly tabulates future ballots.

Not really an easy attack but still a possible one, and these machines have not established a strong reputation for protection against something like a firmware implant

This state uses and retains paper ballots, so it would be yet more difficult to design an attack that would withstand an audit of the paper ballots. But some states don't...

> Even with electronic voting machines it is rare to get enough privacy to hide attempting to tamper with the back of the machine like this article discusses.

An electronic voting machine could be tampered with well ahead of voting day. Verification hashes and the like are only aggregates that are open to trickery on the way to the humans checking them, the ground truth is impossible for a human to process. We cannot sense the state of a computer without its help. In contrast to this, the initial emptiness of a physical ballot box is very much in the realm of human senses, it's perfectly safe to leave them unattended until voting starts.

Because i don't think you are going to just not count people's votes if the entire voting district is believed to have been compromised.

I don't know what the procedure/ or law would be but being that they know who voted i would think there would be some sort of attempt to allow them to revote.

Because attacks on physical ballot collection doesn't scale as well as on electronic ones.

It has to do with the cost. It is more costly to attack physical ballot collection, and riskier.