At 9 a.m. GMT this morning, funding closed on an entity called The DAO. It’s a blockchain-enabled financial vehicle that’s structured kind of like a cross between Kickstarter and a venture capital fund and which now runs autonomously—no humans needed—on the fledgling Ethereum network. The DAO (short for decentralized autonomous organization) raised over US $150 million worth of the bitcoin-like cryptocurrency, Ether, during a feverish, 27-day sale.
The DAO’s launch is feat that should surely stand out as a feather in the cap for the Ethereum network, as it is the most successful crowdfunding campaign yet documented anywhere, ever.
But yesterday, just hours before The DAO was scheduled to open for business and begin taking project proposals, three blockchain researchers published an article outlining multiple flaws in the governance structure of the organization that they say could be used as vectors for attack. The researchers are asking everyone involved with The DAO to temporarily halt funding activities and fix the critical problems.
“The attacks are quite real. So, somebody has to do something about them,” says Emin Gun Sirer, one of the authors of the article and of the blog where it was first published.
The DAO is the first iteration on the Ethereum network of an idea that has been floating around the crytpocurrency space for a few years now, which is that you could take all the functions of an investment vehicle—fund storage, project vetting and approval, fund disbursement, and profit allocation—and handle it on a blockchain, thereby creating what is effectively a corporation without jurisdictional anchors. Equally attractive to some is the fact that a blockchain-enabled organization is completely transparent and does not rely on a managerial class with high salaries to complete its functions. Everything is done by the code, which anyone can see and audit.
What investors who jump on board do rely on, however, is the expertise of the people who write and audit the code. They have to trust not only that the software is secure but also that the governance models work the way they are intended.
This second part is where Sirer and his co-authors, Vlad Zamfir and Dino Mark, say the DAO creators have failed.
Here’s a brief explanation of how The DAO is supposed to work. It’s first created as a contract written into an address on the Ethereum blockchain. The code for the contract specifies all the rules of the game. This was done by a few well-known people in the Ethereum community.
In order to play the game, you send Ether (the native currency on the Ethereum network) to the contract address and you get tokens back in exchange. These tokens signify your proportional ownership over the mass of Ether poured into the contract.
That period just ended. Now, in order to unlock the funds people will present project proposals and the DAO owners will vote on whether the projects are worthy of investment. For example, the same people who wrote the DAO contract are also planning to solicit investments from the organization to fund Slock.it, a project that is hell bent on decentralizing the sharing economy and replicating corporations like Uber and AirBnb as user-owned entities.
At first the voting sounds simple. But there are a few notable details that complicate any game theory analysis of the governance structure.
- Voting is not a DAO participant’s only power. If I have DAO tokens, I can also decide to split from the larger DAO and create my own smaller one.
- I can also sell my DAO tokens to anyone who will buy them.
- If I vote on a proposal, I lose my right to split and I don’t get it back until the polls have closed. Nor can I sell my tokens while voting is in progess.
- In order for a vote to count, a quorum must be reached. The size of the quorum depends on the amount of funds requested in the prposal.
- There actually is a managerial class with very limited duties. There are 11 so-called “curators” who read proposals and vet them for basic flaws and scamminess. They also manage the status of the payment addresses on the funding proposals. In order for an address to recieve funding it must be whitelisted by the 5 out of 11 of the curators.
- The DAO can vote to fire and replace curators.
It’s starting to sound a bit more complicated, isn’t it? I could go on. But the point here is that the voting apparatus has a lot of moving parts. According to Sirer and his colleagues, the machine has not been properly tuned to get the desired outcome.
“ In general what you really want in any kind of a voting-governed structure like the DAO is you want the voters to vote their true preferences. You want them voting in line with what they want to see happen,” says Sirer. In other words, if a token-holder thinks that the proposal will yeild profits and increase the net worth of the DAO, he should vote yes. If not, he should vote no. But that’s not what we’re likely to see, according to the analysis.
“For a number of reasons it turns out that the mechanism encoded in the DAO is not in line with these principles. Certain people have incentives to behave in a strategic fashion,” says Sirer.
For example,Vlad Zamfir, one of the co-authors points to a strong incentive not to cast negative votes in the DAO. Anyone who votes on a proposal also loses the right to split apart from the DAO until the voting ends and the project in question is either discarded or funded. Zamfir argues that this amounts to a cost on no votes which increases the likelihood that people who would other wise vote no and stop a proposal from going through will instead wait out the vote and just split from the DAO if it doesn’t go the way they wanted. In this scenario, the yes voters get what they wanted. The people who were paying attention and disagree at least get to jump ship. It’s the people who didn’t vote and didn’t pay attention who lose the most, who are tugged along into bad projects, potentially ones that have been intentionally designed to profit only a fraction of the DAO owners.
“The people who don’t participate, the people who are just in it for the ride, who are non-active members of The DAO, they’re going to be the ones who get screwed by biases and vulnerabilities,” says Zamfir. “It’s the passive people, who are expecting this to go well because they trust Slock.it and the curators. But instead, the DAO as implemented today may just spend everyone’s money.”
The pro-yes voting bias is one of seven potentially critical scenarios that the authors outline in their paper. At the end they include options for how to fix each problem.
In order to move on a fix, The DAO would have to vote to write new code into a new Ethereum address and migrate all the funds. This would, of course, take time, which is the reason for the moratorium.
If a moratorium does takes hold, it will be most relevant to the Slock.it group, which has been drumming up support for a proposal that Stephan Tual, the COO of the company, says will request millions of dollars from the DAO.
In an interview on Thursday, Tual downplayed the severity of the DAO vulnerabilities. Regarding the voting bias, he said, “ First of all it’s not in the realm of technical attacks because a technical attack would be—we broke your math and we can take stuff out of the contract. This is in the realm of social attacks. But who’s the attacker in this case? This is more a case of the governance model could be improved. Well, duh. Of course it could be improved. It will be improved and that’s the whole point.”
Tual argued that the DAO, regardless of the unexpected participation levels, is still an experiment and that, even more importantly, its fate is no longer in the hands of the people who created the code, but the people who hold the tokens.
Perhaps in concession to a growing chorus of concerned participants, Slock.it has outlined a proposal to the DAO to fund a permenant security team. But Tual says that the group will also go ahead with it’s originally planned proposal.
“We’ll see. It’s just a proposal. Anyone can go and make another proposal. That’s the beauty of the free market,” says Tual. “ If we felt that there was a huge problem that we considered might happened, we would be the first to say “whoops, let’s do something about it. Let’s just address it. Let’s handle it.” But in this particular case, this is more like improvements than anything else , ” he says.
If the curators chose not to whitelist the Ethereum addresses referenced in the funding proposals, then they can shut down the DAO until they are satisfied that the problems are fixed (although the DAO could always retaliate by firing them). This is what Gun, Zamfir and Mark argue is justified and are now pushing for.
“Basically, if there’s any whitelisting or proposals before the DAO changes code, then I will be very concerned. I think the current code has some pretty clear biases and problems with it,” says Zamfir.