Compromise Bitcoin for just $1.2M | Hacker News Reader

Compromise Bitcoin for just $1.2M

matthall2000.tumblr.com

29 pointsmatt200049 comments

This may seem cheap, but there are already many, many orders out there for ASIC, and all of the factories are backed up. Butterfly Labs only recently released models that have been on order for many months (I want to say greater than 6). If you put down $1.2 million, it could be months before you got your machines and by then you would no longer control 50% of the mining.

This dollar amount is also particularly low because ASIC technology is new - most people who have put money into ASICs aren't even mining yet. Once the technology is settled in, controlling the block chain will become more expensive.

Furthermore, the number of things that you can do while controlling the block chain is actually really limited. You still can't spend other people's money. You can prevent people from spending money, but you can't make them spend money. You can double spend yourself, but if the bitcoin community was aware that someone was manipulating the block chain, they would be much more careful about accepting transactions from new wallets, and would reject all transactions from a wallet they knew was controlled by the double spender.

Furthermore, if you did take control of the block chain I personally would dump a few thousand into miners myself, to help regain control of the system. I'm sure that I'm not alone, and the act of bitcoin users simply 'fighting back' may be enough to minimize your control of the market.

And finally, as other people have stated, you can always change the hashing algorithm. Most people use 1 bitcoin client. In fact, this client once had an update that caused an error and forked the block chain and allowed at least one person to double spend $10,000. In the event of a major crisis, there would most be enough bitcoin users willing to fork the block chain that you could indeed get a new hashing algorithm designed to be incompatible with the attackers hardware. The choice is between that and watch your 'distributed' currency fall under the control of a tyrant.

Someone taking control of the mining process IS a risk, and there are some powerful things you can do with that (like mine 100% of all the new bitcoins, taking control of the supply, and double spending, and rejecting transactions by others), but it would probably take a lot more than $1.2 million dollars because people would fight back, and you are still at risk of the rest of the community forking away from your control. That said, you could still do terrible damage and the price would probably plunge, and you may be able to double spend millions of dollars before enough people noticed to start rejecting your transactions. (are there even millions of dollars worth of things you can buy? and would you have to worry about a government getting involved because you committed financial crimes?)

And even if you manage to maintain control, all that will happen is people will stop using bitcoin until you let up. It's much like a DDOS. It takes power (electricity) to maintain that much computation, and the longer you maintain control, the less bitcoin will be worth.

Edit: I want to add that the scariest attacks only happen at 50% control. At 40%, you can only double spend -sometimes-, and I don't think that you would be able to block transactions at all. Furthermore an organized network (and there is much debate about how organized bitcoin could get, after all it is designed to be distributed) could undo any double spending and you would be limited to slowing bitcoin down. At 20% market power, the probability of you achieving a double spend or undoing some transactions is very small.

I agree it would take a long time and be unlikely for a single ASIC-purchasing party to reach 50%. Another organization, an ASIC-datacenter-hoster, could do it.

BFL has an "ASIC hosting program" where those purchasing more powerful BFL machines can put their machines in an affiliate datacenter. There is real incentive to have one's ASICs hosted there, especially given that power requirements are 6-7 times originally forecast (one needs commercial space to run these now), and that this new hardware has an unknown failure rate and a real tangible cost to not working. The hosting center provides direct maintenance from BFL personnel, so one's machine shouldn't be down for more than a day or two for any failure. I wouldn't be surprised if a majority of BFL ASIC purchasers opt to host their hardware in this datacenter.

The hosting program could lead to the mining pool (the default option is for your BFL ASIC to join the mining pool) at the datacenter having > 50%, under complete control of the affiliate datacenter.

https://news.ycombinator.com/item?id=5584783

I don't expect this to happen, ASICs will be pretty widely distributed before anyone can pull this off. However there is a contingency plan for when this happens: switch the hashing algorithm (for example to scrypt, or something memory hard w/ lots of random flow control). This puts all ASICs out of the game at once. Miners will not like it, but if the other alternative is a non-functioning currency, they will likely cooperate.

> switch the hashing algorithm

OH boy now this is interesting -- to an outsider like me, explain who would decide this? Some central committee? The source-code "custodian"? Will BTC fork at that point so every holder can "vote with their (pardon) dollar" but would also need to bet on everyone else's choice?

So it would be a combination of the coders, the miners, and the community. Yes, this would definitely require a hard fork.

Changes to the Bitcoin protocol can be ratified by a majority of miners, in the same "longest blockchain" method.

This is a common misconception, I think. If you try to model the protocol economically, it's pretty clear that it's a majority of currency holders (what the core developers call an "economic majority"), not a majority of miners that matters in determining whether a protocol change is valid.

Exercise for the reader: determine whether these sets are different in a meaningful way.

You can model all you want, the majority of miners determines what is the next valid blockchain. Even if they hold 0 currency, and even if there were no more bitcoins to be mined that would still be true.

> This is a common misconception, I think. If you try to model the protocol economically, it's pretty clear that it's a majority of currency holders (what the core developers call an "economic majority"), not a majority of miners that matters in determining whether a protocol change is valid.

Actually, its neither, if you really think about it. Currency users (in exchange) are more important than either miners or holders (with the caveat that you need some number > 0 of miners to validate exchanges) to whether a protocol change has economic effect, but once some group adopts a protocol change, what you have is a fork into two separate currencies until a consensus is achieved.

Passive currency holders have little driving force in this; currency users are the main driving force, because wherever its being used is where it will have value. Miners have some force because without some of them, the system collapses.

And among currency users, the ones that matter the most are the ones that accept bitcoin for goods and services, not the ones that spend it. (In the "steady state" those should be approximately the same, but as long as mining is still producing coins you can have miners/spenders that aren't accepters, and even in the steady state you could have people who inherit hoards and slowly deplete them as users that aren't accepting.)

No, I mean in a purely technical sense, the majority of miners are the ones who validate blockchains and the ones who have the power to change the protocol.

In comparison, Amazon EC2 offers GPU clusters for $2.10000/hr [1]. Each cluster contains 2x M2050 Fermi cores [2] which each put out ~80MHash/sec [3].

To get the requisite 64686 GHash/sec, you'd need to spin up 404287 of these things, which costs you $849000/hr. This is of course assuming Amazon has 400000 instances of the GPU compute clusters.

The cheaper alternative is to utilise spot GPU instances, costing only $0.346/hr [4]. At that price, i's $139900/hr or so. I am willing to bet that they do not just have 400000 spare GPU instances laying around unused, however.

All this was purely academic, but it kind of amuses me that for a theoretical $140000/hr, you could hyjack bitcoin as it is currently.

[1] http://aws.amazon.com/ec2/pricing/

[2] http://aws.amazon.com/ec2/instance-types/

[3] https://en.bitcoin.it/wiki/Mining_hardware_comparison

[4] http://aws.amazon.com/ec2/pricing/#spot

"To get the requisite 64686 GHash/sec, you'd need to spin up 404287 of these things, which costs you $849000/hr."

Really, I don't think even Amazon has enough of this hardware. And AFAIK these (GPU clusters) are not virtualized.

Makes me wonder what happens if you have a cost advantage in taking the cheapest PC hardware and plugging high-end GPU cards (beyond ASICs)

Keep in mind that a plan like that would take, at best, half a year to put in place. In that time the amount of ASICs in existence will already be far higher than it is today (I would guess at least three times what it is at the moment.)

3x? More like 300x, I'd say. Once the first few roll out of BFL these things will be coming online as fast as you can say "money press".

That assumes BFL won't just run off with all the money.

If the bar is that low, this does suggest that, contrary to some conspiracy theories, governments aren't really out to destroy Bitcoin, since they could've done so by now if they had really wanted to.

There is another possibility: governments just want to control Bitcoin. The same attack that allows double spending can be used to deny transaction confirmations, which effectively denies transactions from happening at all.

There's other things they could be doing.

Bitcoin isn't intrinsically anonymous, so maybe they're just keeping an eye on anyone not being anonymous.

Wait until someone sells polonium or uranium on Silk Road to see if Tor / Bitcoin stand up.

Correction: bitcoin is only "not anonymous" if the transaction history network contains a contact point with the physical world, that is, if some transaction is associated with a physical good shipment or traditional world identity.

Bitcoin is not inherently anonymous. Users have to do stuff to be anonymous when using Bitcoin. To say otherwise is wrong and dangerous.

(http://www.newscientist.com/blogs/onepercent/2011/07/bitcoin...)

(https://news.ycombinator.com/item?id=2800790)

(https://en.bitcoin.it/wiki/Anonymity)

(http://anonymity-in-bitcoin.blogspot.co.uk/2011/07/bitcoin-i...)

AFAIK you don't need much computing power, you just need to be lucky to be the first one to solve the problem and create a new hash. I case you have lot's of tries even a 0.05% chance every 10 minutes might be enough.

Not sure if i make a mistake here, but it seems to be easy.

0,5 < (n choose 1) * (1-0.0005)^n

0,5 < 1 * (1-0.0005)^n

log(0,5)/log(1-0.0005) < n

1385.95 < n

13860 minutes < 10 days

So a 50% chance in 10 days. If you have a 0.05% of the computing power of the whole network.

I'm not sure there's $1.2M worth of ASIC hardware made yet. Also, for that amount of money you could do a lot of malicious things, even to traditional banks.

"I'm not sure there's $1.2M worth of ASIC hardware made yet."

I am pretty sure a large government could make its own ASICs...

You don't think the NSA could free up a few machines and have a couple of Terahash/Sec available in a weekend?

ASICminer is a company producing their own, and from what I've read they certainly have the capacity to overwhelm the network if they wanted.

There is probably no business case for such an operation. To benefit from an extreme price crash you either have to short bitcoin or buy a ton of stuff with it. In the latter case it is hard to imagine a merchant honoring such orders if he is not convinced the blockchain is sound.

As for shorting bitcoin... It's certainly not impossible. But shorting a few Million of Dollars of Bitcoin? If this is done in any kind of trading account (without the money being physically in the hand of the attacker), the trading institution would probably be out of business because of the price drop anyway, before paying the profits to the attacker...

For everyone else... You shouldn't have more than 20 BTC on hand anyway. Preferably a lot less, depending on the rest of your portfolio. In these cases any price drop isn't going to wipe you out and the network resumes normal operation.

See https://news.ycombinator.com/item?id=5585166 for information on BFL's hosting program. It's likely that BFL and affiliate datacenter personnel will control > 50% of the bitcoin hash rate at some point. Here's my comment on that thread ...

BFL has affiliate data centers where those purchasing the more powerful BFL machines can have their systems hosted, for a fee. The default option: "Your hosted units will be added to a mining farm and you will be paid out regularly based on their collective output." Datacenter and BFL personnel will be monitoring the machines for defects and maintaining them. The hosting option makes sense for new untested hardware like this ...

So, it's likely the BFL datacenter mining pool will control > 50% of the bitcoin hashrate at some point.

And of course you forgot that the difficulty would increase?

Who sells ASICS 66GH for $1250? That page is outdated.

Avalons are going for 72BTC that's over $9000

Well, the difficulty increase wouldn't stop an attacker having >50% computing power. Only switching the algorithm would (assuming the attacker would use ASICs)

This would not disrupt Bitcoin for too long. I'd just move the blockchain to a different algorithm and BOOM your ASICs are worthless and Bitcoin will still go on.

$1.2M will shutdown Bitcoin for, at most, a couple days... and then you're left with a bunch of useless ASICs.

And all that assuming that nobody else will get ASICs, which will make the attack more expensive.

It's just not worth it.

Don't you have to get all other bitcoin users to also migrate over to this new algorithm?

And even if it's just down for a couple of days, that's terrible. Bitcoin is still pretty small in the grand scheme of things, but imagine if regular money stopped working for a couple of days.

Miners WILL switch (for their own benefit). Something like this has happened before[1] (though the mining algorithm didn't change) and the community did cooperate. Some miners will be happy about it, specially the ones with GPUs which will be thrown off the game once ASICs arrive (and will get back in the game once ASICs are killed).

Well, if I can get a government to spend $1.2 M just to shutdown a currency for a couple days, I guess we won (specially hardware manufacturers).

Also: Bitcoin will never replace regular money, and it isn't meant to.

IMHO this attack will not kill Bitcoin, only ASICs.

[1] http://bitcoin.org/chainfork.html

The algorithm was not replaced, then it wouldn't be BitCoin anymore.

https://en.bitcoin.it/wiki/List_of_alternative_cryptocurrenc...

It will still be Bitcoin, with a different algorithm... Bitcoin is not just a protocol and some algorithms behind it.

Bitcoin is a community and a currency. Your Bitcoins will still be called Bitcoin, you'll have the same wallet and the same number of BTC in your accounts. The webpage will still be www.bitcoin.org and we'll talk in #bitcoin as we used to do.

Changing Bitcoin to scrypt will not magically turn Bitcoin into Litecoin.

This scenario has been envisioned before. Nothing new here. Bitcoin will prevail.

I think the fact that Bitcoin is a community is its main problem. Miners have acted altruistically a few times in the past, but as the pool grows, getting them to do anything beyond the Nash Equilibrium will get progressively harder.

Bitcoin may prevail, but it will have to go through a painful 2.0 evolution at some point.

Economic majority will ALWAYS win by definition.

And a smart agent with enough resources can lead the economic majority like lemmings down a cliff once it reaches a certain size and becomes dumb.

While your miners are mostly hobbyists that you can access via Skype, it's easy to say "hey guys, let's stop this chain and switch to that other one." Once you have large companies with dedicated ASIC racks in the game, and any change that could impact the short-term bottom line has to go through four levels of management, you are SOL.

Bitcoin has had several value crashing events. It takes less than a few days for Bitcoin to go from $X to $Y (where Y is very much less than X).

These events don't kill bitcoin, but they do make uptake among regular people slower.

> It's just not worth it

Governments are not rational. Look at any government and it's easy to find them doing something very expensive and very stupid. I agree that it's unlikely, but "crypto" usually implies "rigorous".

> but they do make uptake among regular people slower.

Which is not a problem at all! In fact it's good and a sign of bitcoin's success: a free market is self-regulated.

> Governments are not rational.

Well, even if they succeed killing bitcoin (which I doubt) bitcoin won by making a government spend millions of dollars (bear in mind $1.2M is the current price even before ASICs got mainstream) and a lot of time/effort killing a currency that will probably just migrate to an alternate blockchain in a couple days without much effort.

That's assuming you could get your hands on ASIC hardware. None of it has shipped to customers yet, and once it has, the difficulty will skyrocket.

It has but only that first batch. The author also makes the mistake only that first batch sold at they rate the new batch is priced at 75btc.

Interesting to know how much it would cost to build your own machines only reasonable solution at this time.

> Interesting to know how much it would cost to build your own machines only reasonable solution at this time.

Finger in the air calculation: $3-4M. Assumes knowledgeable designers, and good ops people with decent offshore assembly experience.

None of this is particularly difficult from the standpoint of a VLSI. The issue is that profitability is questionable: If you're successful, you pretty much destroy the economic niche that forms your customers.

Whoops: s/VLSI/VLSI designer/

Avalon has had units shipped for quite some time. They're up to their second and third batches now.

If you wanted to damage/degrade BTC with $1.2 million I think the "smart money" wouldn't be investing in the custom hardware to compromise it - but instead to flood the markets with liquidity. BTC is still so thinly traded that a million dumped in pieces could have a serious destabilizing affect. Not that reducing price was the actual subject of the topic but...

I think longest blockchain attacks only apply to double-spending, so you still can't "compromise" coins that are not your own

You can also prevent transactions from being confirmed.

If a few agents start double-spending, that creates a strong disincentive for anyone to accept bitcoin at all, which can doom the entire currency.

Assuming nobody else buys any of this hardware too. That's how I see this anyway. All of these theories assume that nobody else is going to try and do the same...

Reply on news.ycombinator.com