Hacker News Reader

Comment by Strom | Hacker News Reader

You're choosing a suboptimal attack spot in your paper voting attack scenario. Instead of manipulating the thousands of people around the country who count & send out the results of those booths, it's easier to manipulate the far fewer people who do the aggregation for final results.

Also a computer voting system that is exploitable via "one hack" is inexcusably badly designed. Yes hacks happen & adversaries have zero-days, but it's possible to build a computer voting system that is highly resistant to those scenarios. As a very simple example: have a standardized vote format that will be counted by three independently implemented systems, each on a different OS, running in parallel. A great system will go to far greater lengths for protection, but this should already show that a simple "one hack" wouldn't work.

View on HN

It's still a 3-point of failure versus a thousand in a paper vote.

I've seen the British election many times, there are hundreds of people in the counting hall for every district. They count, then they shuffle the ballots around and have different people recount. It would take the bribery of thousands upon thousands of people. People who are under constant scrutiny, not just by those around them, but on TV also.

A computer is almost trivially suborned in comparison. Even our most important pieces of security software, such as openssl and others, get hacked from time to time. If such hugely important - massively used, massively tested - open-source projects get hacked, why on earth would you want to risk computer voting?

It's just standard programmer NIH hubris; we think we can do it better, therefore we should.

>It's just standard programmer NIH hubris; we think we can do it better, therefore we should.

It's not NIH at all. Electronic voting has not been implemented sufficiently. It could be done. It's not about doing it better, it's about doing it at all. It's not-invented-anywhere hubris, and that's plenty of reason to do something.

People aren't much harder to manipulate than machines, and they're also just as transparent. There is no fundamental reason to believe computers can't be more secure than humans when it comes to collecting data and tallying it.

To be done sufficiently, you need to not only prove that your system is fair, reliable, scalable, tamperproof, anonymous etc., but also that you system actually runs. That combination is nigh impossible. That is the fundamental reason, bootstrapping that trust. Imagine you have a 100% open source system, cryptographically verified etc. How do I know that is actually what is running in the voting box?

Secure multiparty computation.

You trust that the right software runs, because unless multiple mutually distrusting parties runs the EXACT SAME software in the same configuration, nothing will happen at all - and they'll hopefully never agree to collude on running anything but the officially agreed upon version.

https://roamingaroundatrandom.wordpress.com/2014/06/16/an-mp...

So who do you trust enough to be distrusting parties? This to me seems to be the same problem as making it proof of work: if the incentives are big enough, the scheme collapses. Look at the creativity that lead to caging and other forms of voter suppression

Not sure. I did consider the risk of bureaucracies becoming cemented within the participating organizations, leading to a harmful drift in values.

But you really only need each department to live for as long as the vote is active, so you can dismantle them and reassemble them afterwards each time. Hopefully that would prevent the establishment of dangerous practices and attitudes.

Whoever is managing it must be held accountable, and must work in full transparency.

Contesting political parties. That's exactly how it's done in many countries already.

I've counted votes a couple times before and can tell you that where I come from in the US recounting is minimal (if done at all). We pair up with a person of our choosing and then count and recount. Collusion is trivial. Just make sure the total counts add up and that your results aren't so wildly disproportionate from the other people's that they'd get reexamined when all the pairs' numbers are summed.

I don't know what you mean by scrutiny on the TV - are local polls counted live on television in the UK?

Yep - this is the kind of thing:

https://www.youtube.com/watch?v=XydeHaqxBbg

It's almost impossible to imagine that someone could bribe that many people to influence an election in the UK. There are 650 seats with over 100k population per seat. I just can't imagine it being bribable on a large scale. Maybe some tiny constituency in a tiny district might get influenced, but the absolute smallest is about 20k votes. That's very difficult to rig.

But you don't have to bribe 100k people. You just have to bribe/influence 50,001 people (and I know that's not exactly true, this is based on a US two party system, not the variety of parties in the UK, but hang with me here.) And in reality, it's as the saying goes for Baseball. "Every team is going to win a third of it's games, and every team is going to lose a third of it's games, it's what you do with the last third that makes a difference."

Every party is going to have a large number of supporters and a large number of detractors, it's what you do with the unaffiliated votes that matters.

Now I'm really going to talk out of my ass, in a horrible, gross generalization with no support to back me up; I'd guess that for many elections, it's less than 20% of voters who actually decide the fate of their country/district. And since we're talking about voters (not citizens or residents/people), we're then talking about an even smaller group of people you must influence. That's why bribery, voter intimidation and pure old fashioned marketing matter; You don't need to get everyone on board, only a much smaller percentage of key voters. Isn't democracy great?

Some UK seats are marginal. You only need to manipulate a few dozen votes to influence those seats.

That's much easier to do.

> it's easier to manipulate the far fewer people who do the aggregation for final results.

It is but parties have their watchers spread around the country to report counts to their central. Any substantial difference will be noted and end up in a recount.

What saves us is the mistrust between parties. With internet voting watching each other would be more difficult. Actually I believe that there will be no internet voting where parties really don't trust each other. There will be internet voting were parties are very naive (almost impossible for politicians) or they somewhat agree not to rig the elections (again, too naive to believe?) or to decide them in advance.

Interestingly enough mutual distrust between the organizing parties can create trust trough cryptography, in secure multiparty computation schemes.

Still no good idea to do remote voting (social reasons), but electronic voting can be made secure. I linked my own scheme before;

https://roamingaroundatrandom.wordpress.com/2014/06/16/an-mp...

> [...] it's easier to manipulate the far fewer people who do the aggregation for final results.

That problem can easily be solved with transparency. In Germany I can do my own tally of the election results as every state reports the votes down the individual polling station. For example, here are the results for Berlin from the last Bundestags election: https://www.wahlen-berlin.de/wahlen/BU2013/ErgebWahllokale/w...

I can also observe the count at my local polling station. They are required to announce their results orally after they're done counting and I can confirm that this is the result reported online.

I've been an election helper and I've always checked that the results online match those that we reported. I never did my own tally, but if the results are close I'm sure someone does them.

Reply on news.ycombinator.com