Going from Bad to Worse: From Internet Voting to Blockchain Voting

hi everyone this is charles hoskinson broadcasting live from warm sunny colorado i just got back from wyoming i was running around gillette wyoming hanging out with the parents taking a look at some ranches and it's just a lovely time this weekend and somebody over twitter pinged me and said hey there's this lovely paper from mit written by ron revist and niha and a bunch of other fellows michael spector and sunoon park mit and harvard talking about blockchain-based voting and the crux of the paper is that internet internet voting bad and blockchain voting even worse and it's actually a really fun paper to read it's about 24 pages long and it has embedded within it a lot of interesting questions and a good framework to look at elections and though though i disagree vehemently with the paper's context although it is kind of ballsy to stand up before the r in rsa there is this is a great example of the power of the scientific method in the academic process it is actually a really good paper to think about elections in general and i'd highly recommend reading it and it provides a pretty good definition of evidence-based elections and they say these are necessary but not necessarily sufficient conditions to satisfy the ability to have auditability and auditing in your election so they kind of outline five items in the paper of thing properties elections should have the secrecy of the ballot software independence voter verifiable ballots contestability and then obviously some mechanism of auditing and then they go on to talk about some of the challenges that both permissioned and permissionless blockchains have in e-voting in general and they take a position that basically online voting and blockchain based voting doesn't really have these properties as a standalone system if it's hybridized with a paper system and you're using the blockchain for a a facility known as a public bulletin board or an authenticated bulletin board then they're not quite as dismissive but it is a very interesting paper nonetheless and i think at the the very end of it and i don't see here where was that section on page security mechanism design critical questions in page 4 16 where they actually go through a bunch of things like stakeholders and adversaries security objectives and the threat model security mechanism design evidence-based elections vetting and transparency and so forth so the problem when you have cryptographers talk about something is that they only have part of the puzzle and you have to really think around the social components of the system as well for example one of the core things that they fight very aggressively in from a security perspective is this concept of ballot secrecy why because one you don't want people to know why you voted one way or the other your vote should be secret but then also you shouldn't be able personally to prove that you voted for a candidate a or b because their primary concern there is vote selling and here's the reality we don't live in the 19th century anymore where vote selling is rampant we live in a highly digitized age and there are far better means for people to interfere and intervene in elections than going to alice and say if you vote for candidate x i'll give you a boatload of money here's a good example united states so if you look at demographic statistics certain demographics tend to vote for certain political persuasions much more so than others for example united states african americans historically vote around 80 to 90 percent democrat in the presidential election so a rational game theoretic strategy if you wanted to buy votes instead of trying to spend five million dollars trying to convince random people to vote for your democrat candidate or pick your favorite would be instead to spend that five million dollars going into places where people typically don't vote but are of that demographic and just simply register them because that eight to nine out of ten you register will likely vote for your candidate so it's a significantly more rational strategy to just spend money this way and technically speaking that's not a violation of ballot secrecy or vote selling or these things so these are examples that whether you're on a paper-based or a blockchain based system you can't really do much about and those are far more effective means of getting hundreds of thousands of extra votes one way or the other okay so these are examples of things that are not addressed in the paper but are incredibly important when you think about the meta of a voting system in general is that if certain things about how certain demographics vote you can just use that fact to take advantage of the system as it is now the other thing is that and this is a fair there's two very fair complaints in the paper one is that we suck at public key private key management the guy who invented it ron is probably a really good person to tell you that and pki is an important thing and the reality is you cannot do e-voting unless you have an authenticated bulletin board it's not good enough just to say i can click a button and vote you need some sort of a symmetry in the system where what you input into the system cannot be forged and has a degree of non-reputability about it the only way you really can do that effectively is with a digital signature at some capacity okay but they very rightly point out that pki is real hard in practice and they say hey hundreds of millions of dollars worth of cryptocurrencies have been lost because we're not so good at keeping our private keys secure that said our entire industry wakes up every day and tries to figure out ways to make this more usable and more secure and so forth they also point some things out in the paper that i think are very unfair i took some notes for example when they said hey let me find the the notes i have here oh like for example they cherry picked this thing as saying that online voting may not increase turnout that's a relatively counter-intuitive thing because anytime you make voting more accessible you'd expect more people to do that and they point out some of the cantonal elections that were done in switzerland that's one of the citations as well as decreased voter turnout in belgium and some things in estonia and canada or how online voting will simply favor one class of people or the other i think most people agree that a hybrid voting system makes a lot of sense and i don't see a reality where you move completely from a paper-based system to a blockchain based system what you would rather do is you would say okay you can do either or for example united states you can vote in person or you can get a mail in ballot okay so similarly you can vote in person with a mail-in ballot or you can do an online vote it's not clear to me how that would disenfranchise voters or reduce turnout the other thing is that you can have with an electronic voting system significantly more involved ballot complexity or faster polling of people so you can use different voting systems like ranked order voting and other things much more comfortably and you can have much larger ballots and more frequent elections and so forth there was a bit of a cherry-picking too when they pointed out that a blockchain-based voting system in moscow was was horrifically broken i happen to have been at the conference where that was discussed i believe that was real world crypto in colombia earlier this year and i remember that presentation quite well and yes it's very fair to point out that that system was terribly broken but that's not necessarily a representative sample of the state of the art and it's very easy for me if i'm trying to pick on a new crypto system to just simply pick on the worst possible systems or systems that are not well constructed there's also a little thing that says governments may also provide legal recourse for victims when they were talking about the differences between online shopping and online banking and how these are lower thresholds than e-voting they seem to make this assertion that there's no means to make voters whole again if there's a compromised election that's just not true it's just materially not true and the reason being is that we have the election but the people elected do not take office until much later after the election for example the united states we had the election in november and the president will not take office until january i believe it's 20th or 21st i can't remember the date which gives plenty of time for auditing oversight contesting and that's one of the properties they push is this notion of contestability so if there's a suspicion that the system upon audit mandatory or enforced is compromised in some way then there's plenty of opportunity to redo the election or plenty of opportunity to invalidate part of the election and and intervene so you can build in to the latency between when the new candidates take office and when the election happened all kinds of recourse so i don't think that's fair at all there's also this concept this is reading verbatim for the paper users of bitcoin other cryptocurrencies have lost hundreds of millions of dollars due to theft fraud or mistake cryptocurrencies have fewer risk absorption mechanisms than traditional banking that's also not true you can build all kinds of things with smart contracts and governance for reversibility oversight you can use multi-sig for things you have escrow accounts there's all kinds of staging mechanisms you put in so you can rewrite the logic of the system so these systems are much more flexible in general than a normal traditional e-commerce system or online banking system and you're not at the mercy of people but okay and so then they they go into things like software independence and i agree that's an incredibly difficult problem you have this issue that we don't have secure operating environments we're looking for them and we do build them in certain domains like for example dealing with classified information skiffs and other things exist and people invest billions of dollars into constructing air gap very sophisticated systems that are immune to foreign coercion or tampering but in general it is very expensive and hard to build a system that is ubiquitously secure this is the holy grail of information security and it very rightfully said that it's difficult to avoid a compromised infrastructure and there's many vectors of attack there so that's why a hybrid system makes a lot of sense because you have means to look at that now voter verifiability this is one that i think is very unfair they say even if about before ballots casting a voter composing about it must be able to verify for herself that her prepared ballot reflects her intended choices i have no way with a paper ballot of knowing if my ballot was actually counted and it actually reflected my choices or not i have to have just faith in external counters that that indeed was counted so let's be clear here that ballot construction is a problem with a digital system what can happen is when i submit my ballot to a public bulletin board i can take a hash of what i've submitted and all that metadata and data and sign that with a did or some sort of identifier and i can commit that to an authenticated bulletin board now if i do that that i can always verify that my ballot indeed was posted and recorded properly now they say well you shouldn't do this because if i can reverse the hash then i can prove to an outside party that i voted for trump or for biden or other things like that and then i could potentially sell my vote so you could create a system where you have plausible deniability where i can verify that i voted a certain way but you can put in another way of doing the verification so that you can make it appear that you voted for someone else so if anyone's attempting to buy my vote i could always make for example it it appears as if i voted for a particular candidate so that's one way you can look at that and you can build a back door into the system where there's an additional piece of information that the government could possess that would allow the government to know which one of those is legitimate but you couldn't prove that now the advantage there is criterion number five auditability now if i have the ability to reverse a hash what the government can do is pick a statistically significant subset of the people who voted and then simply say if you prove that your pre-images are correct and you check that and it's not a mistake in the system we can give you a benefit a tax credit for example so why not have the people who voted themselves actually provide auditing and oversight of the system if you have dids or did aliases cannot connect it to non-reputable commitments in the system it seems like that would actually be a reasonable way of doing things and you actually have a great degree of auditability there as opposed to the existing auditability where if we lose for example the envelopes it's impossible to verify if a ballot is legitimate for male imbalance but okay contestability is another example of that i have to kind of be able to prove i voted a particular way to contest that my vote either wasn't recorded and if you have public bulletin board with pre-image with commitments that are signed by a did alias you would be able to contest if the pre-image wasn't correct because you'd have your pre-image and people would have to forge your signature to to get a different commitment but okay and then they they talk about different categories of voting systems inside the paper and i think that's pretty interesting but they keep just saying over and over again the conclusion of the paper they do talk about end-to-end verifiability in [Music] system level attacks yeah in section 2.4 on page number nine and i was actually very nice that they they cited ari jewels work and the civitas proposal with clarkson and also the u.s vote foundation's nose notes on end and verifiability and there's a great paper actually where they talk about many of the challenges and maturing that's required there so it's quite good and they talk a lot about transparency in section 2.