Epicenter - Learn about Crypto, Blockchain, Ethereum, Bitcoin and Distributed Technologies - Emin Gün Sirer & Vlad Zamfir: On a Rocky DAO

Episode Date: June 6, 2016

Raising $150m+ through its toke sale, ‘The DAO’ has become the most notable decentralized application to date. The ambitious goal of the project is to form an decentralized organizations that efficiently makes investment decisions and generates a return for the token holders. Computer Science professor Emin Gün Sirer and researcher Vlad Zamfir joined us to discuss the various security issues with the daring project and why they’ve called for a temporary moratorium on funding proposals. Topics covered in this episode: How the DAO works What the role of the curators is What splits are and how they became a way to withdraw funds Why the DAO has a bias towards approving proposals How attackers could ‘stalk’ token holders when withdrawing their funds How the DAO can be upgraded Episode links: A Call for a Temporary Moratorium on The DAO Blog Post A Call for a Temporary Moratorium on The DAO Full Paper PDF EB132 – Stephan Tual: Building A Universal Sharing Network On Ethereum And A $150M DAO Understanding the DAO Accounting The DAO website Meher's 'split DAO' reward token attack This episode is hosted by Brian Fabian Crain and Meher Roy. Show notes and listening options: epicenter.tv/134

This is Epicenter Bitcoin episode 134 with guests Vlad Samfair and Emin Gung Cyr. This episode of Epicenter Bitcoin is brought you by Haid.me. Protect yourself against hackers and safeguard your identity online with a first-class VPN. Go to hi.combe slash Epicenter and sign up for a free account today. Welcome to Epicenter Bitcoin, the show which talks about the technology, projects, and startups driving decentralization in the global blockchain revolution. My name is Brian Fabian Crane. And I'm a hero.

Today our topic is very interesting. We are going to talk about the DAO and some of the security challenges that were discovered with its design. Joining us to walk through this topic are Vlad Zamfir and Emin Goun Seer. Both of them wrote a paper pointing out several security flaws in the DAO and calling for a moratorium on proposals right now. So Professor Seer is an associate professor. of Computer Science and Cornell University and Blad is a researcher with the Ethereum Foundation

who works on Casper and the blockchain and their blockchain sharding solution. So before we start, let's have a small intro from the two of them, starting with Goun. Hi, thank you very much for having me here. Let's see, I'm an associate professor at Cornell University in the computer science department. I work on building self-organizing systems in general, And I've also worked on all sorts of things related to cryptocurrencies going as far back as 2002. You might know me from such epicenter Bitcoin episodes as selfish mining. I was one of the two people who discovered the selfish mining attack on Bitcoin. And since then, I've done quite a bit on trying to make the Bitcoin ecosystem healthier better.

This includes building the new thing that we call covenants. and the Bitcoin vaults for securing money at rest, Bitcoin coins at rest. I've also worked on Bitcoin Next Generation, which is a new protocol that is based entirely on the same exact code, sort of underlying infrastructure as Bitcoin, but uses a trick to get around the block size limitation so that it can achieve latencies for one confirmation that are as low as on the order of a few seconds, and throughputs that are limited solely by the underlying network, not by any block size restriction. So I'm also a co-director of the, what is it, IC3,

which is the initiative for cryptocurrencies and smart contracts. Hey, I'm glad. I'm a researcher at Ethereum. I've been working at proof of stake for like over a year and a half now, and I'll start working on blockchain sharding and a few other things here and there. Cool. So how did you guys end up writing this paper pointing out the security flaws in the DAO? Why did you analyze the DAO at all? So it first came to my attention that there might be like some security problems with the DAO when Dino Mark,

had a conversation with Dino Mark, who was like the third author on the paper. And he's the one who kind of led the charge by initially writing a documentation of like how the DAO works and a bunch of attack vectors. And that's like what eventually kind of turned it. into this call for a moratorium. So I was speaking with Dino about it. And then I was also hanging out at Cordell to work on Casper. And Goun and I got to start talking about it. And we realized that it was something

that we wanted to work on in the short term. Yeah, I mean, I think the amazing thing here, too, is just the speed of it just happened. A few weeks ago, this was, well, kind of started at this way for the slotky guys to get some money. And then all of a sudden, it took on this massive life of its own, becoming the biggest crowdfunding campaign ever.

And I was very skeptical, to be honest. But what has been amazing for me to see, and I've only sort of followed it a little bit as an outsider, unlike the three of you, is how much involvement and how a community has arisen sort of instantly to dissect it and criticize it and make other proposals, And it was particularly amazing to see, for example, that original security proposal that the Slokit guys made,

and then immediately detailed analysis of all the things that were wrong with it on blog posts and Reddit and hundreds of comments and discussions. So actually, that made me a lot more optimistic about the project. It's just kind of seeing all the activity that went into that from all kinds of different people. So I entirely agree. So it's a sign of health that these things. are being dissected so quickly by a million eyeballs essentially and it's again a sign of health that people are taking this thing seriously looking at it carefully trying to make it better hopefully we'll come out of this with a much

better doubt that people actually understand and have some assurance in that it implements what they have in mind another thing that I find interesting here worth mentioning is right so there was always this idea of you know what's the dollar like what is the best application for a dollar and where is it going to be used and it's interesting kind of right now that the doll or this first decentralized autonomous organization that's kind of really getting traction and being used actually doesn't really even have so much of a function besides being a decentralized autonomous organization and making some decisions right and that being completely vague what they are.

I mean, okay, there is the presumption that it's going to act a bit like a venture capital fund with some similarities there, although at the same time, right, you could imagine it doing other things as well. So I think that's just a, it's an interesting thing to think about that maybe wasn't so clear, right? I think if you'd asked people before, what is a dog going to be, what's the first successful dog going to be? I don't, not sure if this would have been the one.

Well, so that is true. On the other hand, I'm not so surprised about the turn of events, that the Dow turned meta, and then that one of the first proposals is all about managing the Tao itself, and it seems like substantial effort is going to go into making the Tao better. This is exactly what you would expect with any cool new technology, that the very first steps you take are shaky and uncertain, and people will go into it and try to make it better and much more robust. You're absolutely right that we're in a sort of a funny situation right now.

Dow isn't necessarily funding big money-making proposals being brought in by contractors, but is instead trying to make itself better. And that's a good thing. Absolutely. I think from that perspective is also a positive aspect that they raise so much money because that allows them to kind of do that if they had raised $3 million dollars and they spent the first two million dollars on making the thing itself useful then that would have been a very unattractive thing for investors but you know if

they if they raise on a sixty million dollars they spent ten million dollars or twenty million dollars and making the thing itself actually useful then that could actually work yeah I mean I like that idea but I certainly don't I hope it doesn't cost that much money to make this thing work I mean I hope that like you know with a few relatively easy to implement changes, it would be a lot safer than it is now. And then with a lot more changes, it could be really just like orders of magnitude better than it is now.

But, you know, it's like we have to acknowledge that this thing already is live and we have to kind of, you know, deal with that in the kind of short-term basis. We can't just kind of put everything on hold for like a couple years while. We do a lot of research into like the optimal data architecture. Yeah. So we'll be working. through now on kind of what are these kind of attacks and problems that the two of you have identified with the Dow. But before we start, perhaps it makes sense to just have a broad overview.

Like Brian mentioned that this is meant to be, the Dow is meant to be a sort of a VC firm, right? Something that invests in projects. It's a mix of between the Kickstarter and a VC firm. But assuming that listeners have that basic background, let's kind of go into some of the details of the DAO, like, you know, what is a curator and questions like that. So maybe we should start with just describing the role of the token holders and the curators. What are token holders of the DAO tasks with doing and what are curators of the DAO task with doing? Okay. Well, at a very high level, the DAO is something like a mix between a venture capital firm

and Kickstarter, as you mentioned. So what that means is, like a venture capital firm, it has raised money from the public, and it's going to invest it in projects of different kinds. Those projects come to the Dow in the form of proposals, and they get voted upon. This is where it differs from a traditional venture capital firm, where you would have a professional fund manager make decisions.

Instead, with the Dow, the decisions are being made by the token holders, by participants, investors in the DAO. And these people vote, and their votes carry as much weight as they have tokens in the DAV. So the more you invest in the DAO, the more you have a say on how it spends its money. And the money then is allocated to proposals. The critical difference between Kickstarter and this

is in Kickstarter, you're putting your cash and your cash alone towards a project. In the DAO, you're voting with your share and you're committing other people's money for its projects if the votes are above a certain voting threshold. Okay, and Vlad, what do you think is the role of the curator out here? Can you describe that? So, I mean, at very least, I can describe the mechanism of the curator

that's encoded in the Dow as, like, a contract. So the curator manages a white list of payment addresses to which proposals can be made. So the Dow can only basically, like, send calls and cryptocurrency to addresses on a white list and the curator contract manages that white list. It can add things to the white list and it can remove things from the right white list. So the curator is just this contract that has this ability as far as the Dow that was concerned. But in today's DAO is a 5 of 11 multi-sig that you know has

the ability to add and take away members from the multi-sig and also to change the threshold required. So And then what is the role of the individual curators or part of the multi-sig is something that is still kind of being defined as a little uncertain. And I think that is a whole other rabbit hole that maybe we should say for after we talk about, talk more about how the DAO works and the attacks against the DAO, etc. Okay. So we have walked to like the token holders. So the token holders are people who kind of invested money during the creation phase of the DAO, now hold some tokens and then can vote on different proposals to spend money on certain. projects. Then you have the set of curators which define like high level which addresses

can claim money for for spending on on projects on the DAO so whenever a proposal comes in then it has to first get approved by the curator and then only can it really apply to the Dow and have a have a voting phase in. So allocation of money is decided by voting right. So the final component of the DAO that we might want to work through is the notion of a split. So, Goun, can you explain to us the notion of a split and what it does? Sure. So the purpose of a split is to allow people to take their money out. And the DAO follows a particular, what we would call a pattern in software engineering,

where you don't just get the ether out as soon as you demand it. But instead, the way the split works is the bigger Dow contract splits and creates a child contract. And you become essentially, if everything goes according to plan, you become the sole owner of the smaller Dow. And the rewards that were coming to you in the Mother Dow start going to the child. And in the child, you are the sole curator, and you can do whatever you want. And in fact, if you really want your money out, what you would do in the child is you would put up a proposal to take all the money out to pay yourself,

and then you would approve it because you're the only voter. You have 100% of the votes, and you would then get your money out. It's thus a slightly complicated process. It's not as trivial as just saying, I want my money back. You have to go through the Dow creation phase, and then the proposal phase within the child, to get your money out of the mother DAO. And just to...

Quickly say something about that as well. Initially, the way that, like, Christoph and, like, the socket team conceptualized splitting is as a way to defend against a malicious curator that, like, a lot of the DAO would coordinate to kind of jump to this new, this new DAO that had a different curator. But since then, people have reconceptualized it as being, like, you know, if I don't like what the DAO is up to, I can take my money out. And that's why we've kind of left with this awkward withdrawal mechanism. Because initially it wasn't intended as a withdrawal mechanism and now it's being adopted as one.

Because one of the reasons that I could interpret this as, so maybe I'm wrong here, right? It's because like let's say you invest money in this DAO and initially you should just be able to take the ether back out. But now this has been invested in a variety of different projects that over time will will pay back into the DAO. dollars returns there. So I should have some way presumably, right, if I take my, if I want to take my money out, then still receive the rewards. So is that, that, but that was not the reasoning why you'd have a separate DAW that then these rewards will get redirected to. Yes. So I mean you, that that is like the child, the children DAO can claim rewards from the parent down.

Fourth money they've already spent on proposals in the past. I mean, so, for returns on funds they've spent in the past. Let's take a short break to talk about hide.me. Look, when you're choosing a VPN provider, you want to make sure that your privacy is protected. You know, if a government agency tries to force the VPN provider to hand over some of your traffic or browsing information, will they be able to do that?

And is your payment information attached to the account? These are all things that you want to consider when choosing a VPN provider. With hide.combe, all that's taken care of. For starters, they're based in Malaysia, and Malaysian laws don't require them to keep any logs. In fact, Hyde.Me has no logs of your traffic or browsing history. So even if a government agency was trying to force them to hand over some information, they would be straight out of luck because Hyde.me has nothing to give them.

In addition to that, they use a third-party payment provider, which doesn't give them any of your payment information, so they have no way to link an account to like a credit card or a PayPal account. So even if you're paying with PayPal or credit card, there's no way for Hyde.Me to know which account paid for what. And of course, if you're paying with Bitcoin, then you're completely transparent. And so what we suggest is if you're creating an account with Hyde.me, if you want that extra level of privacy, just make a fake Gmail address and use that to sign in. So that way, you're completely anonymous.

You can give Hitami a try with their free plan. Their free plan includes two gigabytes of data at unthruddled bandwidth. You can use any of their free exit nodes, which are in Alibald. Amsterdam and Singapore and in Montreal and you can sign up for that at height of me slash epicenter. Now if you use R URL and if you decide to go premium down the line it's going to get you 35% off and the premium plan gives you a lot. It gives you unlimited data you can use as much as you want you can connect up to five devices so your whole household fits on the plan and you can use any of their exit nodes all over the world and they've got like 30 of them.

And of course, you can pay with Bitcoin. So give it a try. We would like to thank Haight up Me for their support of WebSenter Bitcoin. So what kind of complications does I add, Vlad, that you said originally the idea of a child's doll was to replace the set of curators, but now that it's being kind of used in a different way? I mean, so it's made it so that like the withdrawal process is complicated

and that you need to start worrying about like this like soccer-ish, which we I guess we should talk about now that it's kind of come up which is the concern that like in your attempts to split alone you might have someone follow you or I'll also end up on the same thing and then you have wouldn't have a majority of the shares so that you couldn't pass proposals so to be to be very clear about this the way the stalker attack works is suppose I want to not have had enough of the doubt I don't want to be part of it anymore and I want to take my money out.

The recommended procedure is to split, so I initiate a split, and normally the expectation was that I would be the sole person who would go for this child DAO where I alone am the curator. But a malicious person can follow me, and he can vote yes on my proposal to split, he can enter my new DAO, my child DAO with me, and now suddenly I'm no longer the, the sort of the owner, the complete controller of the child DAO.

And this stalker that's next to me that's following me around is now in my DAO, and if he comes in with enough funds, he can actually overpower me. He can reverse my decisions. And such a stalker with large enough funds can actually just squat in my DAW, and now I'm at a loss, like what do I do? And there are a bunch of things that have been suggested, The Slokka team has a web page that says the stalking attack is a non-attack. But there are two issues with it.

For one, there is certainly the potential for financial loss. If I split again from the child DAO to a grandchild, then I lose the rewards because the rewards don't carry down through the generations of DAO's. If I leave the child, then I've left it behind as an empty shell. And my rewards that I was getting from the mother DAO into the child, I left them, I abandoned them to my stalker, and he collects a bounty. So that's a big problem. The stalker could also follow me into the grandchild.

And then this is a denial of service attack. Now I'm unable to take my money out. He could actually start to try to ransom me. And so that's a problem. And there are technical tricks that one could play. They're all probabilistic. They all rely on race conditions and timing and so forth. They're very complex, and I think that your typical investor is going to be hard-pressed to pull them off.

And so I don't think that this problem has been addressed sufficiently yet. Leth, go ahead. Yeah, so I mean, I clearly agree with you, Gugin. I just, you know, would like to go in just like a little more detail about, like, what these tricks are. Basically, the idea is that, like, if you can get your money into your Dow or into your split at the last block of the creation period of the Dow, then the attacker won't be able to follow you because they'll have to get in at the next block. And basically to do this, you need to guarantee that you're only going to get in if the attacker hasn't already got it in before you, and you're only going to get in if it's at the last block.

And the attacker also isn't going to get in at the same block as you. So you can make sure of all of these things, especially if you're mining your own block. But it's difficult, and the only way to make it work without, like risking losing your rewards, is to start many, many, many different split proposals and to split many, many different times, but only to follow one of them. And the issue with this is that this is going to cost a lot of gas. So it's going to be costly, and it's going to require writing programs to monitor the blockchain and to inject the right transactions at the right times.

So personally, I think this is a no-go. This is not a reasonable way to build a system. when we build large distributed systems, the kinds of tricks that we're talking about here to defend against a stalker, we call them race conditions. They are dependent on timing, at being there at the right time.

I'm sure almost everybody who listens to this podcast has experienced some race condition or another. Most of the time, my Windows box, I don't have a Windows box anymore, but when I used to run them, most of the time they work okay. On occasion, you get a blue screen. Now, blue screen happens

because the things ended up timing just right for some code to get messed up. So you don't want the safety, the security of your funds to be dependent on these kinds of payment mechanisms. So it's just not good system design. And the other reason why this is kind of a concern is because building in a solution to this would be very, very easy. having a withdrawal function that you could call at any time in order to redeem your tokens for the kind of current share of the ether and also claim future rewards would be a really elegant fix right so so you could have a very actually a very simple way that you should say okay I'm going to withdraw my ether and then have some address that all the other rewards get

redirected through in proportion to what I had and and that's and then that's fixed and if you mean if you don't have that separate step of creating a child to withdraw money yeah then it can happen a lot faster because you don't need to wait like you know seven days to split 27 days for creation and then 14 days for your proposal you could just kind of have it happen that seems like a very straightforward fix so is that something that there's kind of agreement on that that would be a way to solve this and that is that's how it's going to happen or is there all the problems like there's a strong agreement forming around it. I can't say like 100% for sure that that's going to happen, but I haven't heard any,

I haven't heard anyone who like thinks it's a bad idea enough to like defend, defend that, like their claim. Yeah. So I mean, the way we could think of it is like in any system you have, you can have voice and you can have exit, right? Like you're living under the government, you don't like it. You can either speak up or if nothing good happens, you can exit. And it kind of implements this by saying you have voice because as long as you're a token holder you can vote on proposals and allocate money that's your voice and if you don't like the way the system is going then you can exit the problem here is exits are complicated a they take a lot of time as currently defined so they take something like 40 days to execute and then there's also a particular

attack once you try to exit that attack could in theory be some are mounted but it requires a lot of technical profits which most of the Dow investors won't have right so so that that is the current situation with the stocker attack. Today's magic word is moratorium. That's M-O-R-A-T-O-R-I-U-M. Head over to let's-stockbidcom to sign in, enter the magic word and claim your part of the listener reward. So what are the other kinds of attacks you have come across? So as a token holder, what I care about is that I'm able to vote correctly and I'm able to exit correctly if I need, right? So in this case, what we're saying is exiting has become hard and this has reflected itself by having the Dow token price also go below the book value.

So today if you look at Dow tokens, you can buy Dow tokens and if you can exit out of the Dow correctly, you will get back more ether than you put to buy the Dow tokens today. And this is because the exit, the way to exit from the Dow is really complicated, as demonstrated by this attack. The other side is like if I'm a token holder, the other thing I care about is voting. And you've also identified some of some problems with the kind of voting system that has been implemented. So what are these? So the big problem with the voting in the Dow is that it's biased towards yes votes. So think of a fund that wants to make decisions. It needs to capture the true

sentiment of the crowd that it's crowdsourcing from. And to get the true sentiment, what you really want people to do is they want their incentives to be aligned, that if they see positive, expected value in a proposal, they should vote yes without thinking. And if they see clear negative value, they think that this contractor will waste money, then they should vote negatively. They should vote a no. The way it's structured right now, the yes voters have every incentive to vote,

yes, the optimistic happy crowd, that's great. The no voters, the sort of the financially minded careful crowd who think that this thing is going to be a bad idea, they will lose certain rights if they vote no. And in particular, if they vote no, they get trapped. That is, they are unable to split. So you might have the Dow faced with a terrible proposal and the yes voters will vote yes on it.

You're looking at this, you're thinking, this is clearly a terrible proposal. I want none of this. But it would be silly for me to vote no now because if I do, I'm trapped in these idiots if they vote yes, if there are enough idiots around and if most clear-thinking individuals are inactive today or whatever throughout the voting period,

then you end up losing your money to that very proposal. So suddenly the situation is no longer symmetric. And the happy-go-lucky crowd might end up committing you to very bad proposals. And your best bet is to hold off on voting no. So then that creates a secondary issue, which is if you look over the timeline,

you will see yes votes start to come in. And if we're using the DAO as a signaling mechanism, which we are, right? We want this voting to poll the audience, to all the audience and you want to be able to watch as the audience sort of brings in their votes, those votes will come in heavily weighted on the yes side, and then no votes will tend to come in towards the end

if they come at all. So this is a huge issue because it means that the mechanism is not incentive compatible, it's not truthful. And what we call these kinds of voters, we call their behavior strategic. That is not the layman's kind of like, oh, that's clever and strategic.

This is, they're doing. something that is not exactly in line with what they should be doing. And this is bad behavior. It diminishes the Dow's ability to make sound decisions. You're hearing from one crowd, but not the other. And so when you say strategic voting, in particular what it means, right, is that I, as a DAO token holder, it might not be in my interest to vote, for example, against a certain proposal, even though I think it is actually bad for...

you know, the Dow fund and for the returns that the DAO will generate because I have, you know, some other concerns or some things I'm going to try to protect. Absolutely. That's exactly right. You can clearly see that this thing is going to bring negative value, but you're not voting that way. In fact, you choose to sit on the sidelines. You might choose to split at some point if you see the yes votes pile up. The sort of the game theory behind what we will end up seeing is complicated, but we can all agree. that the voting is skewed, that the yes voters, there is no impediment

to their voting, and in no voters there is some structural impediment. We will always hear from them late, and if they're clever, they might not even vote at all and split out. Yeah, and the only thing that I want to add to that is to be clear that it's not just that you give up your ability to split, but you also give up your ability

to transfer your tokens. So you can't sell either, right? And if splitting is kind of maybe not great, because you have to have a week notice, but you maybe would want, so if you think that something might pass and you want to split, you'd have to like kind of speculate

a week before it passes. But for selling, you can potentially sell even a few days before it passes. But, you know, the closer you get, the more of the market will have already reacted to the fact that this thing will pass.

So, you know, it's, you can't split or sell if you vote. And if you think that the DAW is doing great, then you don't mind so much at not being able to split or sell. If you're concerned, then you're more likely to be a no voter, and you're also more likely to want to retain your right to split or sell.

So this dovetails well with a second, I guess we're on number three now. There's another attack that comes from this, which is if you vote on a proposal whose voting period is long. So I ask, I think there is currently a proposal on the Dow that says, do you believe in God? You know, it's like a zero cost proposal. And, you know, you vote yes or no. But whichever way you vote, you're trapped.

This we call a concurrency trap or a concurrent proposal trap. So once I get you to vote on this BS proposal, you're now in the doubt for good until that proposal is resolved. So one vector of attack would be I would put up a zero cost, goofy proposal that lasts a long time. you feel instinctively sort of moved to vote on, oh, I'm an atheist, I hate, da, da, da, da, or, oh, my God, this is my faith, whatever. You know, some dumb thing that gets people riled up and moving, and the moment they cast their vote, they're in for at least the voting period, the end of the voting period.

And now you can start attacking them with other proposals that have shorter durations that come in. So how would that work? What could another, like, let's say, a shorter duration? proposal, how could that be used to attack somebody that's been locked in? Sure. So imagine that you want to get money out of the Dow with like a majority attack or a surprise attack. What you would try to do is get a bunch of people to be locked in so that they can't sell

or split because then you're more likely to be able to have your funds when you, when your proposal gets passed. So if I can lock, to imagine if I could lock down like 50% of the Dow and I have like some number of the token. I mean, then and I can get a yes vote on a proposal, then like some people otherwise would be able to split like can't because they were because they voted on the long, the long silly one. So so that means I get my crazy proposal to pass and now all these 50% that have been locked in because they really wanted to assert their opinion on some stupid thing and you know my dumb proposal passes now they can't get out. So So to the extent that they're invested, right, the proportional share is going to go into my proposal.

That's the scenario here, right? That's very much the scenario. I would like to add one request. If anybody does that after hearing this podcast, they buy us a drink, all four of us. And plus, you know, of course, five of us. Yeah, like these are really interesting attacks. Like, I would also like to make a restatement of the first attack, which is the bias against voting no. Let me just add one note

before you do that. So if you get these people to vote, then they don't actually have an extent of not to vote in your shorter length proposal. And so depending on your threshold, it may or may not be a good idea to bribe them into getting locked

because as soon as they're locked, they don't have a no bias anymore. Yes bias anymore. So, Vlad, I think there are so many attacks against the Dow that some of them cancel each other out. yeah yeah but the very fundamental problem i think here is the bias against uh against vote so so so the way you can imagine bias is like this this this this is the famous famous thing which is called a survivorship bias right a survivorship bias is that uh in the middle ages or

even now they like you find people that said you know i i pray to god I went to see my sip shank, sank, but I was saved because I prayed to God. You know, I prayed to God some disaster happened, and I was saved from this disaster because I prayed to God, right? But you can't see people that are the opposite, that prayed to God, that some disaster happened and ended up dying, because once they are dead, they weren't there to popularize the notion that they prayed to God met with a disaster and died, right? So in society all that you're going to see is people that say I prayed to God and I was saved

The whole like you know the natural system is sort of designed in a way to have us see only these kinds of people Now what what you're saying out here is because a no investor has a cost Right once once he votes no he is locked in the Dow he can't sell his share he can't exit the DAO etc so so we start to see that these kinds of people don't do anything at all. And the signal that comes out from the DAO is always a positive signal. Okay, let's fund this proposal. It gets reported in all of the news media.

Let's say there's a huge proposal, 20 million, and there are lots of yes votes, it gets reported in the news media, it gets covered on Epicenter Bitcoin, it gets all of the publicity behind yes, because everybody seems to be saying yes, and ultimately that proposal passes just because the system is aligned that way.

In reality, though, there might have been many people that don't want for this proposal to pass, right? And that kind of selection bias makes the whole voting system of the Dow itself broken, right? Absolutely. So we would like to have a truthful mechanism for voting that is symmetric on both ends. And that way you can count on people voting their, you know, what's the way. what they perceive and you can count on the outcome of the of the vote to then be representative of what the crowd thinks about the idea any kind of bias like this survivorship bias etc skews that whole thing makes it down into sort of a dumber dumber money that can be lost

easily so with the stalking attack before there was an easy fix or at least it's seems with that withdrawal functionality, is there something like that here, or is this much bigger problem that is hard to solve? It's considerably more complicated here. So there are techniques from game theory for how to solve it, but evolving from where we are now to one of those is not, this is not a two-week patch.

This is, there is some thought to be given. The game theoretic techniques for fixing this, they typically require some secrecy of how things, at some point in the process. And it's hard to achieve that secrecy on a blockchain. So I think there's going to be some research needed to come up with mechanisms that are truthful and are adapted well to blockchains.

And kind of like this problem, this brokenness of, of the voting system itself, this creates a very interesting dynamic today, right? Like ideally what we would want is that the voters, that the DAO members be incentivized to upgrade the DAO right. So we can always think that maybe, okay, the DAO has certain attack vectors against it. Okay, that's fine and dandy. They'll just create a proposal. Somebody will say, okay I'll build a new framework to upgrade the DAO, submit that as a proposal, maybe walk away with like a few hundred thousand dollars, that the DAO will sponsor, and then

this guy can build the next DAO framework, and then all of the token holders can adopt that DAO framework or something like that. So in theory, you could imagine that the DAO itself becomes a principle, and in order to upgrade, it recruits an agent, and then that agent allows the DAO to upgrade itself. But if the voting system, system itself is broken, then we cannot rely on this self-upgradation path of the DAO itself, right? Because the basic thing, voting that's needed for the self-upredation path itself is broken. So that's why this creates like a chicken and egg problem of how do you really upgrade something in which you can't even elicit the opinions of the community members efficiently.

Yeah. So I think this is like, you know, one of the reasons. why a moratorium is appropriate at the moment. But there is a couple of things that we can say. One of them is that proposals that have, you know, spend zero ether are less prone to a yes bias. Although there's still a bias against voting, it just no longer towards yes or no. It just so probably, you know, the participation won't be very high,

but there's not going to be a bias if it spends zero ether. And so we could solicit. the Dow's opinion without actually having a bias. And then in terms of the upgrade path, there's a, there's a, there's a, there's a, there's a, there's a, there's a, there's a, there's a, there's a, three percent of the stakeholders to agree, uh, the Dow will be upgraded to a new contract and all of the, they're all the extra rewards, everything will go over there.

Regarding upgrading the contract, I guess another question here is what's actually the process here because, uh, Right now there's a GitHub account where the code of the dollar is on. Obviously, the GitHub account isn't managed in a decentralized way, right? So they're managed by a few people who control that account, who can control, you know, where the pull requests get accepted there. How does that influence this whole process of upgrading the dollar? I mean, technically it doesn't.

I mean, technically, you know, you can call the upgrade contract, you know, the upgrade contract call doesn't know anything about the GitHub repos. I mean, it just knows about contracts on the blockchain, right? So the question is more, you know, what would it take to get the community support to the point where we have like a, you know, 53% of the tokens agreeing to the upgrade? And I think that's going to take, you know, a lot of clear communication and good education. Because this reminds me a little bit of the kind of situation in Bitcoin, why you can say, okay, the Bitcoin core, you know, this doesn't have necessarily a particular status who controls

those repos. But then it does have actually a very significant importance because that's where people look for and that's where people turn to and you have this kind of, what you call it, a shelling point. And then moving that somewhere else is actually really, really hard. maybe not maybe in a technical level this isn't an issue but on a social level kind of you know coordinating everyone to sort of start uh considering something else i mean maybe there's going to be nice tools developed here for the dao that will make that easier but i mean i think if we look at

bitcoin that has been uh there has been a challenge yeah absolutely i think uh in the case of the dao it's a little better um but uh the sort of the the frontline discussion, the voting and so forth, it's happening at consider. It's what is Dow, that consider it, whatever it is. It's happening on different platforms, and you could, oh, it's also happening on Reddit as well. It's partially decentralized, partially going through channels with, you know, managed by Slackett. But I think, you know, as you said, as we move forward, we're going to build other tools and this channel should be a little better than what we saw with Bitcoin. So it's not possible, for example, with the Dow to have happened the same thing

that happened to, you know, block stream and almost all the core developers. So it is, it is its own thing now. It has started, it's live, it's going, it's ticking forward. So there is no worry that somebody could usurp the Dow, in my opinion, it's broken at the moment. It's limping along, but it's nevertheless its own thing, and it's decentralized, and it's not short of taking over the entire curator set and so forth.

It's not going to be easy to take it over. So it may be limping, but at least it's decentralized. Yeah, but realistically, I think you're right. I think that the cleanest, most elegant way to do any upgrades It's just with stock and 100% on board. And I think that, you know, that will very likely happen. A few weeks ago, we told you about the G-Tech blockchain contest. We asked you to submit your blockchain startup ideas for your chance to win 50,000 euros

in grant money from RWE, G-Tech and Glebombus. Well, over 100 startups submitted their ideas, including 16 of you, our listeners. Well, the results are in, and the winner of the grand prize is Arcade City, a project with a radical idea to cut the middleman out of ride-sharing. And the runner-ups are cargo chain, a blockchain distance to improve international trade, especially in the shipping industry, and Clippers, a decentralized permanent document storage solution intended to guarantee intellectual property without a middleman. Congratulations to the winners, and we wish you lots of success with your projects.

If you have a blockchain startup idea and think Berlin could be the home where you are going to grow your company into a billion-dollar behemoth, and make sure you check out G-Tech, or the G-Tec on-Pont. entrepreneurship center. G-TEC has a lot of programs, workshops, startup academies, provide office space to help companies grow quickly, work on really innovative concepts. So make sure you check out their website, check out G-TEC.Berlin. That's G-T-E-C.B-E-R-L-I-N. And we hope to see some of you in Berlin soon. We would like to thank G-TEC, R-W-E, and Clubbumbus for their support of representative Bitcoin. So one question I have here is like,

there seems to be a principal agent problem out here, right? So if you think about it, before the DAO existed, there was just SLOCIT, right? Now, once the DAO exists though, now, now you have like a principal, which is like the DAO, and then there's an agent which is like Slocut. And Slocut really wants to become that agent because like, they're coming up with all of these proposals that are security related and they might come up with something that's also upgrade related, right? So now you have like two different entities, right?

And because they are like these two different entities and their mutual interests may not be the same anymore, right? Like they might be different, right? and there might be other entities that come up with other upgrade mechanism. So there might be like three or four concurrent proposals that come up, right? So this upgrade mechanism could itself be more complex and lead the doubt to be forked just on how, if there's a discord in the community as to how the upgrade should happen. That might be possible in some sense.

But hang on. The principal agent problem is not as bad as well. one might imagine. So the Dow is what it is. And Sloket is at this point kind of like any other contractor. They have expertise over the code and they propose fixes and now it's up to Sloket to convince the community that they're capable of understanding the attacks on the Dow, admitting them, and then addressing them. And if they can do this to the satisfaction of the community, then they can move the funds over to Dow 1.1 or whatever. or maybe the 2.0 version, which has a slightly different mechanism for soliciting vote.

So I hope that they can do this, but we'll live and see if they're capable of doing this. I'm a little worried that they were blindsided by so many attacks at the game theory level, but I hope that they will quickly find the necessary expertise to fix this. Yeah, and I think there's another principal agent problem that we should kind of talk about, which is basically, and it's not entirely clear what the relationship is, but like, you know, between the curators and the token holders, right? Are the curators meant to serve the interests of the token holders? Are the, you know, is, the curators aren't exactly chosen by the token holders. The curators are, you know, initially they were selected by sloppy, but now they get to, they're kind of self-selecting and they get to, like, use their multi-sake to rotate the curators. So it's not really clear, like, who the boss of the curators are. I mean, they're kind of, like, it's not clear how the curators will be paid for their work either, right? I mean, some people would suggest that the curator get paid by the Dow, in which case the curator would be an agent of the Dow. But it's not clear that that has happened or will happen or whether the curators today won't want that.

So there's definitely, like, a conversation to we had about, like, what is the role of the curator? So like in absence of a salary for the curators, right? So the curators, like the original definition of the curators was, okay, this is like a multi-signature account that is tasked with taking all of the people that want to send proposals to the DAO and like filtering that list to a smaller list of addresses that we think are coming up with genuine proposals, right? It was like meant to be like a filter. So now, but is that a number? right? Is that right you are describing it? So it's interesting because there's lots of different interpretations of this, right? So that's like that's an interpretation that I think is not the most common.

I think there's kind of, there's like a few classes of interpretations. One of them is the one that you gave, definitely not the most popular. The more popular one is curators check write code and identities so that like the people who have proposals are identifiable. If they say like fraud or something, then maybe we could use, we'll be able to like track them down somehow, and then just make sure that the proposals that are being spent are really genuinely to contracts that correspond to the proposals. The payment addresses really correspond to the proposals.

That's like one common view on what the curator should be doing. And then another common view is that the curators should be protecting the Dow against majority attacks. These two things are actually kind of incompatible because if you believe that the curators are only involved in this kind of automatic function, then they can't really really be able to this kind of automatic function, then they can't really safeguard against majority attacks. You know, if you, the question of whether or not it's their job to like, you know, take, select some proposals based on their merit, generally people don't really, don't really believe that that much.

So I have heard a lot of token holders say that they believe that curators are kind of the human stopgap measure in case something goes wrong that we can't predict and that we haven't kind of fortunate, you know, don't already have logic in the doubt to deal with. So like in light of all of these different interpretations, like how does your role get defined, right? Like ultimately you want like some kind of, let's say charter from the community. Yeah, this is what we are tasked with doing, right? How do you go to that point? And then if you're not being paid for it, then how does the incentives of the curator and the DAO ever align?

Yeah. Those are great questions. So I mean, I think definitely we should try to solicit the opinion of token holders to deceive what they believe the curators should do. And then give the curators an opportunity to decide, like, okay, is this something I'm comfortable doing or is this something where I think maybe the legal liability may be too much and I should maybe step out of the monthly sick? So I think, you know, and then I think also, yeah, how the curators get paid will also

be, it will kind of factor into this somewhat. Because, like, you know, if the curators are there to make sure that the Dow earns a higher expected return by like being really careful that only really great proposals get in the white list, then I feel like yeah, like it would make sense for the DAO to pay the curators. It depends on what the role is. And you know, whereas like perhaps if the role is just checking the identity, perhaps it makes sense for the proposal, a person who makes the proposal to pay the data, pay the curator, right? Because it's like on a proposal basis as opposed to, you know, a service for the DAO. So it's how the question gets

resolve about what the role is is going to feed into how the curators get paid. And I think it's just kind of an unfortunate thing that's happened because of, I mean, I'm not like something's exactly clear to me why. I think that some people might have just opened up the curator page and just kind of trusted that the curators have their back because it says like they're down the top, like the curators will safeguard the doubt. I think some people who read the white paper and really involved in the community really think the curator should be doing almost nothing.

But also there's people who are in the community who believe that the curators are this stopgap measure where it's just like the humans in the code in case that something bad goes wrong. So let me chime in a little bit. So the underlying, one of the big underlying issues with the curators is that they be in a position to have as legal as small a legal liability as possible. So to that effect, they are not, this is, I think, in everything that Slokket has produced, it's fairly clear that the curators are not in a position to, should not be in a position to exercise business judgment. They're not supposed to say, this is a stupid idea.

That is supposed to be left up to the crowds. And so that makes sense. And so that's fine. But then, Slokit, when they realized that there were certain attacks possible, they realized the majority attack was possible. And we kind of glossed over this, and we should kind of mention this. In the original DAO paper, the way the curator abstraction comes about is, it gets introduced, is they first talk about the following attack. If there is somebody who owns 53% or more of the DAO, then they can come up with a proposal that says,

I hereby award 100% of all the DAO funds to myself, the 53% holder. And then they vote with their majority, then they take all the cash, and then they go away. And so that leaves behind 47% of the holders in a pretty terrible situation. And Slokkett realized this, and then they introduced the concept of a curator. So at a high level, they did the right thing in some sense, because there is going to be no mechanism that can foresee, no automated mechanism, that can foresee every eventuality. It would be a folly to design a system with no human input at all.

So I agree with their sort of gut sentiment. But they failed miserably, I think, in realizing how the majority attack could be waged. If I'm a 53%er, I would not go out and come up with a single proposal for 100%. If I did, then their defense works just fine the way they envisioned it. But I could just as easily come up with two proposals, one for 50%, the other for another 50%. And now I fly under the radar. And if you think, well, that's also kind of obvious, well, fine. about I divide it down to 10% proposals and then vote with 53% of the vote for each of them.

And, you know, if anybody, I make them sound semi, semi-intelligent and semi-attractive. You can make a deal with a company that's already somewhat looks legit so that you don't need to, you know, you can have a front that it passes a smell test. Right. So the curators at the moment are tasked with the impossible. And this is one of the attacks that we outline in our paper. that just by plaintively looking at the proposals, they cannot tell what's a majority attack, what's not. And so if I may go meta-meta for a second,

I think this also underlies a fundamental problem with the way Sloket has been approaching these problems, these issues. So I'm a professor, right? I teach people. And one thing that I see time and time again when I talk to students and I try to teach them synchronization, this is a difficult topic, requires sort of writing correct code,

that can be invoked at any time, kind of like writing smart contract code. So you identify a problem to a student. And if the student is a green student, they just arrived, they don't know anything, you know, or, you know, they're just a sophomore or whatever. And you say, look, terrible event X can happen. And then they immediately go into this mindset.

We're like, oh, X happens here. I'm going to make X not happen. And how does X happen? Well, X happens because A happens and B happens and C happens. I'm going to put a check. in between B and C. And they think this is a good way to sort of handle a fundamental problem.

And I have to teach them. It takes about a year of real hard work to get people out of this mindset, out of this case-by-case thinking. Because then what you have are these, they go into this voice. It's always a thin voice like this. They start talking like this.

Well, that can't happen because then A would have to happen and B would have to happen. But I put a check there so it can't get to C and therefore X cannot have. This is crazy. This is not how you design secure robust systems. Secure robust systems stem from some global invariant.

You just say X is not going to happen, and I have safeguards for it everywhere pervasively. Not one of these like case-by-case-by-case things. So the curator abstraction, if you ask me, came about because of this case-by-case thinking. Some of the case-by-case thinking, some of the instincts, the gut instincts of the developers are right. You know, when your checks that you have programmed,

are going to fail and they will fail, then you need the human touch at some point. You need to bring that in some fashion or another. But the way it was done here is strange, if you're asking. So your point is that we need something like a curator, but their role would be different or should be different. And do you have a view on what that should look like or where that human intervention should be in the system?

All I can say is it needs to be defined. And Vlad is working really hard to sort of try to define the roles of the role of the curators at the moment. I don't want to do thinking and research on the fly here, but I can certainly come up with different techniques if we sat down to do this. But at the moment, the curators, I believe, are flying blind and they're in a tough situation. When they're faced with a difficult decision, what do I do? Is it my job to comment on this or do I ask the crowd? Which one is more in line with the Tao thinking? Which one is more in line with legal responsibilities?

Even though most of us hate to mention that, there are legal responsibilities that fall on the shoulders of the curators. How do we do this? It's not clear at all. So we need a spec for the function of a curator, a specification. Yeah. And the interesting thing is that, like, you know, different token holders have different opinions about, you know,

that should be and like you know ideally like everyone would be on board on exactly what it is right the token holders and the curators and you know a socket um and i think that um you know one one thing that we should maybe spend a moment thinking about is like what is the power of the curator like what at most could the curator do right so like at very most the curator has complete control over the white list you know the order in which things get whitelisted duration for which things get whitelisted when things get unwhitelisted They have a relatively clear ability to control the order and frequency of proposals. Now, I'm pretty sure that everyone agrees that this isn't what the curators should be doing, especially because the order and frequency of the proposals,

if the Dow wants to spend his money at a certain rate, will determine exactly how the DAO spends money. And I think everyone agrees that that's not the role of the curator. At the moment, like the curators are not at all at a determine. or aren't prepared to determine, like haven't even begun to organize the charter of proposals to the Dow. And I don't think that's something that ever we want to see. But it's important to realize that the curators have a tremendous amount of power. And so the scope that we have in defining the role of the curator is very large.

And I think that, you know, we need to be careful that, like, you know, we don't have the role be so broad that the curators can do anything and be responsible for anything. But also we don't want to be so narrow that the curators can't prevent, you know, these kind of outcomes that I haven't been accounted for. So like, like this, this ties into the next section that you want to go at least like path forward, right? Like, so we have all of these kind of complicated questions that are, that are coming up, right? Like, how does the role of a curator get defined? And the other, the other big question is, like, what do you think happens if, the moratorium succeeds like like you're calling for a moratorium on any proposals to the dao at all now what would you do once it succeeds and what do you think what kind of

avenue of opportunities does the moratorium open and on the other side if it doesn't succeed what kind of things does that open so um you know i think that like today we do de facto have a moratorium because like curators haven't started our white listing yet, although enough of them haven't, like, agreed publicly that they will support the moratorium for it to be, you know, approvably a thing that has happened. But I think that, like, the kind of ideal case would be that, like, we in the, you know, as soon as possible, come out with a fix for some of these security problems that gets, like, the 53% threshold vote to upgrade the whole contract.

And, like, if that happens before any proposals to spend money, I think that would be, like, an amazing outcome. That is like one possible thing, but that would require consensus widely between token holders and, you know, people who understand the Dow about like what the appropriate short-term solution is. And so like that's one potential kind of outcome. You know, but you're right.

Like, hereators could decide that like they don't want to hold a moratorium anymore and they can start right-listing proposals. And then basically the, if they do that and there's still a yes bias, then it might be that they're determining by choosing the order of the proposals how the DAO is spending money. But, you know, so it's hard to say what are all the possible things that could happen. It's not sure to say what we should aim for, which is that like we quickly form a strong consensus about what to do and execute it.

have a 53% quorum in the contract vote for an upgrade? So in my opinion, the road ahead is a little complicated. It's not going to be an easy fix, so we can patch certain things in the short term. There are some immediate, mostly non-controversial things we can do to make the Dow provide more reassurance to people that they can get their funds out when they need to. But in general, the problem of building a good, robust system that can ask the crowd and get healthy responses back is far from a done deal.

It's a complicated topic, and it might well be beyond the reach of, you know, many small teams. So there is a lot of work to do ahead. And I was going to mention even the fact. So, for example, even on a good day, you can have strange behaviors in a system. like the Dow. What's a strange behavior? Well, strategic behaviors, people not voting what they think. This can arise from the fact that there will be multiple concurrent proposals, and people might have preferences like, I want these two proposals to get funded only if they both get funded together.

And so that might cause people to vote in strategic fashion, even when there is no malicious actors, even when everybody wants the best for the system, given their evaluations, given their subjective viewpoint, they might be driven to vote no on proposals that will bring money to the DAF. So trying to curtail such activities is actually really, really difficult. Designing that kind of a mechanism is non-trivial, and I think we'll live and see if the Slocket folks can pull it off or if some other team can come up and propose an alternative suggestion. So just to be clear, I mean, Goudon and I, I think, are in complete agreement.

The kind of narrative that I just gave has just to do with. the short-term fixes rather than the long-term kind of let's make the voting system like you know state-of-the-yard voting system so what are you used on on how long that's going to take to get to a system that's really robust and that goes maybe fixes some of these the yes bias and some of these other more fundamental problem is that a matter of six months a year or is that a much bigger project. I think you need to be a little more specific about what fixes you want, right?

So, I mean, I think we could have a withdrawal function very soon. I think the yes bias will take, you know, so withdraw function or if we yes bias on the order of a small handful of months, but like some of the other things are, I mean, even the yes bias thing and if under certain conditions

is a quite tricky research problem. So, you know, anything more than, anything more than just the withdrawal, I think, is kind of unknown how long it'll take to fix. Entirely agreed with Vlad here. Those are the right timetables. I don't think it's going to take more than a year. But as Vlad mentioned, there are some quick fixes possible, which we should apply very, very fast. And there's some medium term on the order of months kinds of issues that need to be addressed. All right. Well, thanks so much, guys, for joining us. That's really

exciting and I think it's exciting to have this new project where so much activity is going on, so much research and another way of actually looking at what's going on with the Dino is this is kind of the first real large scale attempt to do decentralized governance, which is a topic that we've often asked about in the context of Bitcoin and unfortunately it's hard to do in that context. But this is hopefully going to build a lot of the tools that will power all kinds of of decentralized systems in the future? Absolutely. This is a fantastic experiment.

So let's not lose sight of the bigger picture here. No such thing has ever been attempted. We have never seen anything of this kind of this magnitude before. We've never seen any financial instrument with these particular properties before. It's decentralized. It has a life of its own. When we wrote the paper, we were working against a deadline imposed on us, not by any person, but by a computer program.

We had to finish that paper and get it out the door. It was, by the way, it was one of the worst papers I have ever written in terms of, you know, the amount of, in terms of presentation and so forth. For it to be an academic publication,

it needs much more. But we had to go and publish it because the deadline was being imposed by this weird entity called the Dow. You couldn't negotiate with it. You couldn't push it back. So it's really odd. These things have now entered our legal

system. It's going to be a lot of fun for a bunch of lawyers to figure out how to integrate these with what the institutions we currently have. There's this question of, you know, if corporations are people, then what is the Dow? There are a bunch of other exciting things coming up. So this is going to be a fantastic field for a whole lot of people, for technologists especially. This is a fantastic new opportunity. So let's not forget that while we talk about the tax and source, you know, forth. The attacks are important, they're significant. But we kind of understand the process by which we address, we can address them. And at least some of us are making efforts to try to move in that direction. And overall, the big picture is this is fantastic. It's just so

amazingly new, amazing features are embodied here that the opportunities are really, really exciting. Yeah, I mean, I completely agree. I think it's super exciting and super fun and super interesting and you know I think that like a lot of people are having a really fun time kind of like how cyphorpunk this is I mean this is like you know super exciting by all like standards in terms of like you know people who are into like decentralization and kind of peer to peer crypto systems for managing like real stuff cool excellent well thanks so much for coming on guys this was great having you on again and I'm sure I'm sure that will not have been the last time.

Thanks, Brian. Thank you, Brian. Thank you, Mayor. Thanks, Mayor. It's been really nice. Yeah, so Episode of Bitcoin is part of the LTV Network. You can find this show and many other shows on Let's Talk Bitcoin.com. We put out new episodes every Monday.

You can get those through your favorite podcast app or watch the videos on YouTube.com slash Episandemiccoin. And we are still running this t-shirt contest. So if you leave us iTunes Review, just email, show with epsenopaccom and we will send you a t-shirt so thanks so much and we look forward to being back next week