how secure is the Bitcoin-network against a 51% attack at a point in time?

[BNO]

Hi folks,

i would like to find a measure for how the secure network is at a certain point of time. My dream would be that there is a blockchain.info like site that calculates once a day or so how much it would cost in $/btc to rent the needed hashing power from lets say amazon aws to perform the attack. Of course it should assume that the attacker is smart and chooses the least costly option to achieve 51%.

If there isn't something like i think it would be really cool for the bitcoin-community to have a comprehensive one number measure of security. It would help in discussions about blocksize and future discussions about network security, number of nodes, the whole Fee-market debate a LOT if one would have this chart.

Because besides talk about new cool services and the ever question of who satoshi is and the price of bitcoin the rest of the discussion are a very high percentage: network security, blocksize, number of nodes, mining profitability etc. And all those would be easiert to talk about if there was such a measure...

If there isn't such a site, i would still be interested in how the number is for todays network.

[1715224994]

1715224994

[1715224994]

1715224994

[1715224994]

1715224994

[1715224994]

1715224994

[1715224994]

1715224994

[21coin]

I don't think all of amazon AWS or any cloud service for that matter can actually perform a 51% attack, if it does come it will come from the collaboration of the top three hashing pools. Cloud services simply can't deliver that much of hashing power.
As for how to prevent it, there are plenty of threads here about that, either google or rummage through the mining and bitcoin discussion.

[BNO]

I don't think all of amazon AWS or any cloud service for that matter can actually perform a 51% attack, if it does come it will come from the collaboration of the top three hashing pools. Cloud services simply can't deliver that much of hashing power.
As for how to prevent it, there are plenty of threads here about that, either google or rummage through the mining and bitcoin discussion.

I think you can rent on amazon machines with a decent graphic card- so one could calculate out of this how much it would cost for example to have this hashing power for e.g. 24h I would be interested in the calcualtion of this number. Unfortunately i'm lacking the knowledge to do it myself. I'm not interested that an insider (pool, hosted mining whatever) turns evil. That would be another discussion, just the number for renting you hashing power to attack. For that purpose i wouldn't care if AWS actually has the amount (or whatever service you would think would be best to perform attack), just pretend it was there.

[Amph]

I don't think all of amazon AWS or any cloud service for that matter can actually perform a 51% attack, if it does come it will come from the collaboration of the top three hashing pools. Cloud services simply can't deliver that much of hashing power.
As for how to prevent it, there are plenty of threads here about that, either google or rummage through the mining and bitcoin discussion.

I think you can rent on amazon machines with a decent graphic card- so one could calculate out of this how much it would cost for example to have this hashing power for e.g. 24h I would be interested in the calcualtion of this number. Unfortunately i'm lacking the knowledge to do it myself. I'm not interested that an insider (pool, hosted mining whatever) turns evil. That would be another discussion, just the number for renting you hashing power to attack. For that purpose i wouldn't care if AWS actually has the amount (or whatever service you would think would be best to perform attack), just pretend it was there.

i doubt any amount of gpu can cut it, based on the old 7970, you can have something like 2 MB per card right now, with the best gpu available, maybe 3MB(i didn't follow how strong they are on sha256, lately), you need 200 peta to do an attack

200 peta are 200B mega, which mean 100B gpu (basically 14gpu for each person on this planet), none has all those gpu combined, not to mention the wattage, not attainable by a long shot

[LiteCoinGuy]

there was a website with this data but it is offline / broken 

http://www.coinometrics.com/bitcoin/brix

[adamstgBit]

AWS doesn't have bitcoin mining hardware, so you'd be mining with CPU power

so expect to max out AWS's available servers and get less than 1% hashing power.

[DannyHamilton]

GPU is useless for attempting a 51% attack.  AWS doesn't have enough GPU available for you to mount such an attack even if you rented every GPU they have available.  And if you did try to do that, they would raise their prices for GPU processing to meet the huge demand that you would be placing.  As such, the cost would increase exponentially and you still wouldn't have 51%.  If you want to perform a 51% attack, you are going to have to gain control of enough ASIC hash power.

Keep in mind that if miners are making significantly more than they are spending, then there is an incentive for more people to get involved in mining and to spend more on mining to get a larger portion of the available profit.  The additional people involved means that the revenue is spread thinner until it is no longer profitable for people to get involved.  This has an effect of reducing the profits of miners until "they are just barely earning more than they are spending".

Also keep in mind that if miners are mining for profit, and they are spending more than they are making, then they will quit mining (since they aren't making a profit).  This reduction in people involved means that the revenue is consolidated among fewer people until it becomes profitable for the remaining miners to stay involved.  This has an effect of increasing the profits of miners until "they are just barely earning more than they are spending".

So, if we assume that the vast majority of miners are mining for profit, and that they are just barely earning more than they are spending, then it would (incorrectly) seem like this isn't too difficult to estimate.

The average daily sum of the revenue of ALL mining in the world is approximately 3625 BTC.  The current (at the time I am writing this post) exchange rate with USD is approximately $231.  This means that the global revenue expressed in USD is approximately:
3625 BTC/day * 231 USD/BTC = $837375 USD/day

Since we are assuming that the vast majority of miners are "just barely earning more than they are spending", we are assuming that there is just a bit less than $837375 USD/day being spent on hash power.

There are 2 ways that an attacker could instantly control more than 50% of the global hash power:

1. They could buy up half of the currently operating hash power (if they could find enough existing miners willing to rent or sell that hash power for the exact same amount of money as they are already earning).  This would reduce the global hash power that they attacker doesn't control to half of the $837375 USD/day, and would cost the attacker half of $837375 USD/day.  Therefore, it would cost the attacker about $418688 per day.

2. They could add to the network the same amount of new hash power that already exists.  In this case the amount of hash power on the network would double, and the attacker would have half of it.  Since the current network is spending $837375 per day, it would cost the attacker about $837375 per day.

Unfortunately, these calculations are based on an attacker instantly having half of the global hash power and being able to very quickly acquire that immense amount of hash power at the same cost as purchasing a very small amount of hash power.  In the real world, there are several other factors in play.

Trying to buy up such a large amount of existing hash power that is already mining for a profit would create a very large demand in an environment where there isn't enough supply.  Basic economics makes it pretty clear that this huge demand against a relatively slow growing supply would result in skyrocketing costs for the attacker to buy incrementally more hash power.  They may be able to buy the first 2% or 3% at current costs, but the next 5% might cost 10X as much, and the next 5% after that might cost 100X as much.  As they found less and less hash power for sale, they might have to offer a lot of money to convince any of the remaining controllers of hash power to give up that control.

Trying to buy up such a large amount of new hash power would run into the same supply&demand problem.  There just isn't enough new hash power manufactured for the attacker to be able to quickly buy or rent as much hash power as the network is currently using.  Buying up what little is available would encourage the manufacturers to increase their prices.

Meanwhile, if the attacker was willing to wait for some time to gain control rather than having it all happen in an instant, then they could just add enough hash power to the network to make it unprofitable for others to mine.  The attacker would be taking a loss in their mining, just like everyone else, but this attacker isn't motivated by mining for profit. They are motivated by gaining control of more than 50% of the hash power.  Therefore, they would run enough extra hash power to make it unprofitable and then wait for the most unprofitable miners to quit mining.  As the network stabilized, they would again add more new hash power, and wait for more miners to quit mining.  Initially it might just cost them 5% or 10% of the current global revenue (somewhere around  $62800 per day), but they would be gaining revenue from the blocks they mine which would offset the costs. It might take them months (or even years) of mining at a slight loss before they could drive out enough competition to be controlling 50% of the network hash power.  Since it is difficult to predict exactly how much of a loss per day they would need to take to drive out other miners, and it is difficult to predict exactly how long it would take for enough miners to quit, it is extremely difficult to predict with any accuracy the total amount it would end up costing them.

On top of all of this there is an ever changing exchange rate which can encourage or deter new miners and additional hash power that the attacker doesn't control.

[21coin]

GPU is useless for attempting a 51% attack.  AWS doesn't have enough GPU available for you to mount such an attack even if you rented every GPU they have available.  And if you did try to do that, they would raise their prices for GPU processing to meet the huge demand that you would be placing.  As such, the cost would increase exponentially and you still wouldn't have 51%.  If you want to perform a 51% attack, you are going to have to gain control of enough ASIC hash power.

Keep in mind that if miners are making significantly more than they are spending, then there is an incentive for more people to get involved in mining and to spend more on mining to get a larger portion of the available profit.  The additional people involved means that the revenue is spread thinner until it is no longer profitable for people to get involved.  This has an effect of reducing the profits of miners until "they are just barely earning more than they are spending".

Also keep in mind that if miners are mining for profit, and they are spending more than they are making, then they will quit mining (since they aren't making a profit).  This reduction in people involved means that the revenue is consolidated among fewer people until it becomes profitable for the remaining miners to stay involved.  This has an effect of increasing the profits of miners until "they are just barely earning more than they are spending".

So, if we assume that the vast majority of miners are mining for profit, and that they are just barely earning more than they are spending, then it would (incorrectly) seem like this isn't too difficult to estimate.

The average daily sum of the revenue of ALL mining in the world is approximately 3625 BTC.  The current (at the time I am writing this post) exchange rate with USD is approximately $231.  This means that the global revenue expressed in USD is approximately:
3625 BTC/day * 231 USD/BTC = $837375 USD/day

Since we are assuming that the vast majority of miners are "just barely earning more than they are spending", we are assuming that there is just a bit less than $837375 USD/day being spent on hash power.

There are 2 ways that an attacker could instantly control more than 50% of the global hash power:

1. They could buy up half of the currently operating hash power (if they could find enough existing miners willing to rent or sell that hash power for the exact same amount of money as they are already earning).  This would reduce the global hash power that they attacker doesn't control to half of the $837375 USD/day, and would cost the attacker half of $837375 USD/day.  Therefore, it would cost the attacker about $418688 per day.

2. They could add to the network the same amount of new hash power that already exists.  In this case the amount of hash power on the network would double, and the attacker would have half of it.  Since the current network is spending $837375 per day, it would cost the attacker about $837375 per day.

Unfortunately, these calculations are based on an attacker instantly having half of the global hash power and being able to very quickly acquire that immense amount of hash power at the same cost as purchasing a very small amount of hash power.  In the real world, there are several other factors in play.

Trying to buy up such a large amount of existing hash power that is already mining for a profit would create a very large demand in an environment where there isn't enough supply.  Basic economics makes it pretty clear that this huge demand against a relatively slow growing supply would result in skyrocketing costs for the attacker to buy incrementally more hash power.  They may be able to buy the first 2% or 3% at current costs, but the next 5% might cost 10X as much, and the next 5% after that might cost 100X as much.  As they found less and less hash power for sale, they might have to offer a lot of money to convince any of the remaining controllers of hash power to give up that control.

Trying to buy up such a large amount of new hash power would run into the same supply&demand problem.  There just isn't enough new hash power manufactured for the attacker to be able to quickly buy or rent as much hash power as the network is currently using.  Buying up what little is available would encourage the manufacturers to increase their prices.

Meanwhile, if the attacker was willing to wait for some time to gain control rather than having it all happen in an instant, then they could just add enough hash power to the network to make it unprofitable for others to mine.  The attacker would be taking a loss in their mining, just like everyone else, but this attacker isn't motivated by mining for profit. They are motivated by gaining control of more than 50% of the hash power.  Therefore, they would run enough extra hash power to make it unprofitable and then wait for the most unprofitable miners to quit mining.  As the network stabilized, they would again add more new hash power, and wait for more miners to quit mining.  Initially it might just cost them 5% or 10% of the current global revenue (somewhere around  $62800 per day), but they would be gaining revenue from the blocks they mine which would offset the costs. It might take them months (or even years) of mining at a slight loss before they could drive out enough competition to be controlling 50% of the network hash power.  Since it is difficult to predict exactly how much of a loss per day they would need to take to drive out other miners, and it is difficult to predict exactly how long it would take for enough miners to quit, it is extremely difficult to predict with any accuracy the total amount it would end up costing them.

On top of all of this there is an ever changing exchange rate which can encourage or deter new miners and additional hash power that the attacker doesn't control.

Wooh that was quite a post, too ADD to go through it all. Care to tell whats the price(only) for conducting the attack again(without the long statements)

[BNO]

Hi Litecoinguy!
   

there was a website with this data but it is offline / broken  Lips sealed

http://www.coinometrics.com/bitcoin/brix

Thanks for trying!

[spazzdla]

Hi folks,

i would like to find a measure for how the secure network is at a certain point of time. My dream would be that there is a blockchain.info like site that calculates once a day or so how much it would cost in $/btc to rent the needed hashing power from lets say amazon aws to perform the attack. Of course it should assume that the attacker is smart and chooses the least costly option to achieve 51%.

If there isn't something like i think it would be really cool for the bitcoin-community to have a comprehensive one number measure of security. It would help in discussions about blocksize and future discussions about network security, number of nodes, the whole Fee-market debate a LOT if one would have this chart.

Because besides talk about new cool services and the ever question of who satoshi is and the price of bitcoin the rest of the discussion are a very high percentage: network security, blocksize, number of nodes, mining profitability etc. And all those would be easiert to talk about if there was such a measure...

If there isn't such a site, i would still be interested in how the number is for todays network.

................ Even if you took all of the computing power of the ENTIRE FUCKING PLANET you wouldn't hit 1% of the hashing rate.

ASCI >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> super computers >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>everything else.

One would have to make or buy ASIC's to attack the network.

[BNO]

Hi Danny,

thx for that long of an answer! But i must say i don't agree.

Since we are assuming that the vast majority of miners are "just barely earning more than they are spending", we are assuming that there is just a bit less than $837375 USD/day being spent on hash power.

I really want to check what it would cost to rent this. The reason for that is that one can not calculate this as you are doing it, because this number is much to low. This is because after a miner invests in mining equipment the assumption is this: he can not use it for anything than mining, so he can leave that costblock out of the equation and he will continue as long the mined bitcoins are worht more than the electricity he needs to create them. You can believe me on this one. In economics this is called sunk cost and the problem of average cost (which do not apply here) and marginal cost (here you say the money for the hardware is already lost fuck it, i keep mining as long 1BTC is worth more than the electricity needed to create one).

To the second half of your post: Yes i know that this number would be theoretical and in practise the cost for the hardware would go up if you quickly wanted to buy rent that amounts. But i would rather have just this number and then one xould say: "theoretical X Dollars" and as a second layer of security but actually doing that would be much more expensive due to demand and supply. But i would like to keep this simple and stupid because speculating about demand and supply of Asics calculating power is in my opinion not usefull here.

I was just throwing in AWS is there maybe a service for renting ASICs. Are asics used for something else than bitcoin mining?

[spazzdla]

21 if you don't have 5 mintues to read that be ready to get owned at life.

[DannyHamilton]

- snip -
Care to tell whats the price(only) for conducting the attack again(without the long statements)

Sure.  Here's the TL;DR:

- snip -
it is extremely difficult to predict with any accuracy the total amount it would end up costing them.
- snip -

If they wanted to instantly control more than 50% of the global hash power it would absolutely cost more than $418688 per day and would quite likely cost millions per day.

If they were willing to wait, it might cost much less per day (depending on what the exchange rate does during that time), but it would take months or years of continuous effort to build up to more than 50%.

[BNO]

................ Even if you took all of the computing power of the ENTIRE FUCKING PLANET you wouldn't hit 1% of the hashing rate.

That can not be true since the bitcoin network is on planet earth...

And if you meant that planet earth - bitcoin network even than i have my doubts. Someone you can bild this:
http://www.google.de/imgres?imgurl=http://static4.businessinsider.com/image/51b20dd9eab8eaa874000001/heres-the-2-billion-facility-where-the-nsa-stores-and-analyzes-your-communications.jpg&imgrefurl=http://www.businessinsider.com/pictures-of-the-nsas-utah-data-center-2013-6&h=1608&w=2146&tbnid=38Tl8fhPqfRqjM:&tbnh=94&tbnw=126&usg=__pm3iSrbvJNHkVk9AdbEgTAxp-nI=&docid=mNRjxJkZQW4iZM&sa=X&ved=0CDUQ9QEwAmoVChMIhrCah6r-xwIVRSxyCh1lYwhk

Maybe has 10 or 20 billion for some Asics too....

[21coin]

21 if you don't have 5 mintues to read that be ready to get owned at life.

Do I have to if someone can simplify it for me?
And I have some ADD to take care of, you'll find out how difficult it is reading when you have it.

[spazzdla]

................ Even if you took all of the computing power of the ENTIRE FUCKING PLANET you wouldn't hit 1% of the hashing rate.

That can not be true since the bitcoin network is on planet earth...

And if you meant that planet earth - bitcoin network even than i have my doubts. Someone you can bild this:
http://www.google.de/imgres?imgurl=http://static4.businessinsider.com/image/51b20dd9eab8eaa874000001/heres-the-2-billion-facility-where-the-nsa-stores-and-analyzes-your-communications.jpg&imgrefurl=http://www.businessinsider.com/pictures-of-the-nsas-utah-data-center-2013-6&h=1608&w=2146&tbnid=38Tl8fhPqfRqjM:&tbnh=94&tbnw=126&usg=__pm3iSrbvJNHkVk9AdbEgTAxp-nI=&docid=mNRjxJkZQW4iZM&sa=X&ved=0CDUQ9QEwAmoVChMIhrCah6r-xwIVRSxyCh1lYwhk

Maybe has 10 or 20 billion for some Asics too....

LMAO, just no, that's all you get.

[spazzdla]

21 if you don't have 5 mintues to read that be ready to get owned at life.

Do I have to if someone can simplify it for me?
And I have some ADD to take care of, you'll find out how difficult it is reading when you have it.

Apparently I have add... so did every kid in my class...


Anyways, basically the ONLY way to compete with the bitcoin network is with ASIC's. Application specific intergrated circuits.  If it isn't built for the purpose of hashing it's hashing ability will be a joke.

Even that NSA facility buddy thinks could take down bitcoin, it wouldn't even register on the network if it started hashing.


Buying the ASIC's to do this is almost impossible as the price of the ASIC's would skyrocket as they were sold.
One threat is some govs team up, make their own ASIC and use them to decimate the bitcoin network.  I have large doubts about their ability to pull that off.  It's also probably going to piss tax payers off the govs decided to use their money to decimate bitcoin just because.

[21coin]

21 if you don't have 5 mintues to read that be ready to get owned at life.

Do I have to if someone can simplify it for me?
And I have some ADD to take care of, you'll find out how difficult it is reading when you have it.

Apparently I have add... so did every kid in my class...

LMAO now that are the simplifies text I'm talking about, no offense danny.

[spazzdla]

21 if you don't have 5 mintues to read that be ready to get owned at life.

Do I have to if someone can simplify it for me?
And I have some ADD to take care of, you'll find out how difficult it is reading when you have it.

Apparently I have add... so did every kid in my class...

LMAO now that are the simplifies text I'm talking about, no offense danny.

I don't buy excuses. Give my summery a read.

[21coin]

21 if you don't have 5 mintues to read that be ready to get owned at life.

Do I have to if someone can simplify it for me?
And I have some ADD to take care of, you'll find out how difficult it is reading when you have it.

Apparently I have add... so did every kid in my class...

LMAO now that are the simplifies text I'm talking about, no offense danny.

I don't buy excuses. Give my summery a read.

Got that down already, tell me something new, and have a look at my first post, I said nearly the smae thing there.

[BNO]

Isn't there a way to put a number on it. As much as i appreciate the effort of Danny this is not how one can calculate it. If no one knows of a place to rent asics it would be O.K. to make the calculation for buying those. Is here no genius who can calculate what buying the right amount of ASIC's would cost? If yes your input would be much appreciated...

[spazzdla]

IMO it is not possible to put a cost on it because one you start to buy or rent such a large number the cost to buy or rent will SKYROCKET.  Furthermore if people find out what your plans are with the machine they are renting be prepared to watch those machines disapear.

IMO the second those rented machines attempted a 51% attack they'd be shutdown. Wouldn't even be suprised if that was a clause one agrees to when renting.


It's kind of like buyin 1 million bitcoins, at the current price it would cost 230 mill however after you started attempting this the price would start skyrocketing.  The actual cost of getting 1 million BTC is unknown.



Sorry to be a doucher I didn't get your question at first.

[BNO]

Spazzdle and 21coin: Please shut up with this crap. Sorry for you 21coin that you have ADD. But there was some bold text. That contains The answer of Danny.
I think one can NOT calculate it this way but to some it up for you: roughly 420.000 USD /day for renting 50% of the EXISTING network (miners rent him the power) or twice the amount if you buy new gear (the 100%+ odd you buy become than the 50%+ of the new network). So very intelligently he made the distinction of buying half of existing network or buying/renting new gear.  

[BNO]

IMO it is not possible to put a cost on it because one you start to buy or rent such a large number the cost to buy or rent will SKYROCKET.  Furthermore if people find out what your plans are with the machine they are renting be prepared to watch those machines disapear.

For the purpose of this answer just assume:
- It does NOT matter that price will rise - please TOTALLY ignore this
- you buy/rent your hardware outside the netwok
- Absolutely no further speculation/estimation on how the network would respond, what core devs would do or not do or whatever, just a calculation

[spazzdla]

IMO it is not possible to put a cost on it because one you start to buy or rent such a large number the cost to buy or rent will SKYROCKET.  Furthermore if people find out what your plans are with the machine they are renting be prepared to watch those machines disapear.

For the purpose of this answer just assume:
- It does NOT matter that price will rise - please TOTALLY ignore this
- you buy/rent your hardware outside the netwok
- Absolutely no further speculation/estimation on how the network would respond, what core devs would do or not do or whatever, just a calculation

The you really don't want to know the chances of a 51%..

[DannyHamilton]

- snip -
That contains The answer of Danny. I think one can NOT calculate it this way
- snip -

You are mistaken.  If you are going to pay someone to give you control of their hash power, that is exactly how to calculate it.

- snip -
this number is much to low. This is because after a miner invests in mining equipment the assumption is this: he can not use it for anything than mining, so he can leave that costblock out of the equation and he will continue as long the mined bitcoins are worth more than the electricity he needs to create them. You can believe me on this one. In economics this is called sunk cost and the problem of average cost (which do not apply here) and marginal cost (here you say the money for the hardware is already lost fuck it, i keep mining as long 1BTC is worth more than the electricity needed to create one).
- snip -

You are mistaken.  If a miner can get "more than the electricity he needs to create them" by mining with his hash power, and he can get slightly more than that by renting out his hash power, then the profitable thing to do is to rent out the hash power.  Therefore, (since you want to ignore the effect that the increase on demand will have) the attacker should be able to rent the hash power for approximately the same amount as the revenue that is generated.

I can prove this statement, and will do so later in this response.

- snip -
I was just throwing in AWS is there maybe a service for renting ASICs.

There are many "cloudmining" services.  In my opinion, the vast majority of them are ponzi schemes. The few that are not ponzi schemes are probably scams.

Are asics used for something else than bitcoin mining?

ASIC stands for "Application Specific Integrated Circuit". It is a circuit that is custom designed for a specific application.  In the case of bitcoin mining, that specific application is performing the exact proof-of-work that bitcoin requires.  The only other thing that it would be useful for would be other systems that use the exact same proof-of-work that bitcoin uses.  There are a handful of altcoins that use the same proof-of-work, but that's about it.

For the purpose of this answer just assume:
- It does NOT matter that price will rise - please TOTALLY ignore this

Fine.  I think it is a silly thing to ignore, because the number you get will be ludicrously low, but if that's what you want then that's what we'll do.

- you buy/rent your hardware outside the network

This doesn't really exist.  All bitcoin mining equipment that exists is available to be used on the network.  I assume you are suggesting buying newly manufactured equipment that hasn't yet been used to mine bitcoins until the attacker starts using it.  That equipment currently gets bought and connected to the network about as fast as it becomes available, but for your imaginary world we'll go ahead and pretend that an infinite amount of new hardware is available to be rented at whatever the current rent cost would be.

- Absolutely no further speculation/estimation on how the network would respond, what core devs would do or not do or whatever, just a calculation

So, an imaginary number that is entirely meaningless.  Got it.  In that case, the number would be very close to the estimate that I provided earlier: "about $837375 per day"

Now that we got that out of the way, lets look at how I can prove that "the attacker should be able to rent the hash power for approximately the same amount as the revenue that is generated", which means that a minor 51% attack could occur at almost no cost at all (since the revenue generated would offset the costs involved).

If you think about it a bit, you'll realize that every miner that is not solo mining is effectively renting their hash power to the pool operator.  The miner has no control over what is being done with his hash power.  He simply receives a proof-of-work request from the pool operator, and then reports back to the pool operator if/when he completes that proof-of-work.  In return, the pool operator pays the miner for the use of the equipment while it was attempting to complete the proof-of-work.

How much does the pool operator pay for all that hash power that he is "renting"?  Interestingly, he is paying with the bitcoins that he is mining, and generally is not needing to toss in any more of his own money beyond that.  Clearly, the cost that he is paying per hash is "approximately the same amount as the revenue that is generated".

If someone wanted to acquire control of 51% (or more) of the hash power, all they would need to do is set up some pools that pay out a little bit more than any other pool.  Miners will switch their hashing over to the attacker's new pools since they'll get a better profit.  The cost to the attacker will be only slightly more than the revenue generated (since no "honest" pools will be able to pay more than they are generating). As long as the attacker is careful about what they do with their 51%, the miners won't have any reason to switch to any other pool.  Among other things, this will allow the attacker to occasionally orphan blocks from competing pools, and will allow them to occasionally create a double spend transaction.

Ah, but you might say "When/if the miners see what the pool is doing, they can/will switch to some other pool that pays them a bit less, but which is not attacking the network".  But remember you said you wanted to include "Absolutely no further speculation/estimation on how the network would respond".  Furthermore, if you are willing to make an allowance for the possibility that "miners could switch pools", then to be intellectually honest, you need to make an allowance that the services "renting hashpower to the attacker" can stop renting to him if they see what he is doing with that hash power.

[BNO]

I enjoyed reading your post, but thats not what i'm looking for.

Ah, but you might say "When/if the miners see what the pool is doing, they can/will switch to some other pool that pays them a bit less, but which is not attacking the network".  But remember you said you wanted to include "Absolutely no further speculation/estimation on how the network would respond".  Furthermore, if you are willing to make an allowance for the possibility that "miners could switch pools", then to be intellectually honest, you need to make an allowance that the services "renting hashpower to the attacker" can stop renting to him if they see what he is doing with that hash power.

I see where you are coming from here. But i really want hardware outside the network. The reason for this is, that i have some trust in the existing network but zero in the outside. This is an assumption on my side, but not a strong one i think. To say that Bitcoiners in general would not care about politics is absolutely not true. Any bet they are more liberal and less trusting in government than the average citizen. So i really want to keep that assumption.

For the rest: yes the assumptions are not really realistic, that doesn't matter. First i just want the "raw meat", cooking the meal out of it is then the second step. Don't want to mix it all together...

[DannyHamilton]

I've given you numbers that you are asking for and you've rejected them.

I've given you the math behind the numbers, and you've disregarded it.

I've given you real world examples that support the numbers and the math, and you continue to say that you want a different number.

I give up.

If you don't like the number, and you want a bigger imaginary number, please tell us what your imaginary number is, and how you intend to get to that number?

[BNO]

Look Danny when someone asks for a calculation under those assumptions:

For the purpose of this answer just assume:
- It does NOT matter that price will rise - please TOTALLY ignore this
- you buy/rent your hardware outside the netwok
- Absolutely no further speculation/estimation on how the network would respond, what core devs would do or not do or whatever, just a calculation

Than that's what he wants. Presenting a calculation that changes the assumption to: rent the hardware from the inside of the network just violates rule number 2 is simply not an answer. Just see it as an question in university, you would not get any points for calculating this with "your own" assumptions when the question states different assumptions.

But it was still interesting to read your very well laid out explanation about that you could rent a lot from the inside and how to do it. Just not what was asked..

O.K. maybe i'm explaining a bit the reason behind assumption n. 2. The bitcoin network represents a "mix" of different efficient hardware, combined with different electricity cost. The variance within this "group" is so high that makes a calculation basically meaningless if want to look for an outside attacker. Because the outside attacker would start with most efficient hardware, would make an "Factory order" custom desing (that's why assumption with rising prices is not wanted, it would be built up on scratch). And as i wrote if one does believe in the network as a whole it does not make sense to assume people would rent out the nsa their gear... At least that's what i'm interested in..

[GetClams.com]

It would take a unique set of extreme circumstances -


an extreme increase in the price or availability of electricity due to geo-politcal events could potential reduce the hashing power making it possible.

interruption to the internet - http://www.cnn.com/2012/07/10/tech/web/internet-down-eagleman/


So possible but highly unlikely.

[BNO]

@spazzdla: WTF!!! don't know what you are talking about

O.k. i found this page: https://en.bitcoin.it/wiki/Mining_hardware_comparison

and i think the amount needed to buy just the hardware (assuming the next batch has same efficiency) is roughly 145 Million $

[DannyHamilton]

Look Danny when someone asks for a calculation under those assumptions:

For the purpose of this answer just assume:
- It does NOT matter that price will rise - please TOTALLY ignore this
- you buy/rent your hardware outside the network
- Absolutely no further speculation/estimation on how the network would respond, what core devs would do or not do or whatever, just a calculation

Than that's what he wants.

And it's what I gave you.  Multiple times.

See here:

2. They could add to the network the same amount of new hash power that already exists.  In this case the amount of hash power on the network would double, and the attacker would have half of it.  Since the current network is spending $837375 per day, it would cost the attacker about $837375 per day.

Unfortunately, these calculations are based on an attacker instantly having half of the global hash power and being able to very quickly acquire that immense amount of hash power at the same cost as purchasing a very small amount of hash power.

That's what you asked for, right?  That number TOTALLY ignores the fact that the price will rise. It assumes that the hardware will be new hash power that is added to the network, so it meets your "rent your hardware outside the network" requirement. It makes "Absolutely no further speculation/estimation on how the network would respond, what core devs would do or not do or whatever".  It's just a calculation.

Then I state it again here:

I assume you are suggesting buying newly manufactured equipment that hasn't yet been used to mine bitcoins until the attacker starts using it.  That equipment currently gets bought and connected to the network about as fast as it becomes available, but for your imaginary world we'll go ahead and pretend that an infinite amount of new hardware is available to be rented at whatever the current rent cost would be.

- Absolutely no further speculation/estimation on how the network would respond, what core devs would do or not do or whatever, just a calculation

So, an imaginary number that is entirely meaningless.  Got it.  In that case, the number would be very close to the estimate that I provided earlier: "about $837375 per day"

Presenting a calculation that changes the assumption to: rent the hardware from the inside of the network just violates rule number 2 is simply not an answer.

Which is why I added that example at the end, after I first answered your question and used the transition phrase:

Now that we got that out of the way

Just not what was asked..

I'm not sure you even know what you want.  You seem to want someone to just make up a "per hash rental cost" that is big enough to satisfy your desire for a really big number, and then to multiply that by the current hashes per day.  I did that for you and even used a very realistic per hash rental cost, but apparently you want someone to make up a bigger number that doesn't have any justification behind it.  Please let us know how big of a "per hash rental cost" you want it to be, and I'll do the multiplication for you.

[NorrisK]

Besides the enormous cost, they also need to stand to gain something from it right? That would probably mean spending about a couple of million in bitcoin on something to later nullify the transaction. Where can you spend those amounts of bitcoin at the moment?

[BNO]

so in case somebody is interested here is the answer i would come up with.

If the attacker came from inside the US According this site:
http://www.eia.gov/electricity/monthly/epm_table_grapher.cfm?t=epmt_5_6_a

cheapest industrial rate for electricity would be in Washington with 4.5cents/KW/h

then cost would be:

cost for buying 63.000 Antminer : approx 145 Mill $
additional hardware cost for each unit (Powersupply etc.) estimated 200$ then additional hardware cost: ~12.5 Mill $
cost of electricity for 24h (assuming 50% overhead for cooling and operation, 4,5cents/KW/h): 233 000 USD
so if the attacker wants to run 30 days of attack: ~ 7 Mill $ for electricity.

So far ~ 165 Mill USD for Hardware and 1 Month of electricity:

+ Shipping and handling those 63.000 machines
+ Workers for unboxing and setting up machines
+ Rent for industrial warehouse in area with good enough Internetconnection
+ IT Technician overviewing process

[knowhow]

Makes no sense someone invest such ammount to attack bitcoin,they hack emails and try to hack some pc to get acess to bitcoins...why would someone do that if they can get bitcoin sending virus or by other stuffs.

[notiely]

Hi folks,

i would like to find a measure for how the secure network is at a certain point of time. My dream would be that there is a blockchain.info like site that calculates once a day or so how much it would cost in $/btc to rent the needed hashing power from lets say amazon aws to perform the attack. Of course it should assume that the attacker is smart and chooses the least costly option to achieve 51%.

If there isn't something like i think it would be really cool for the bitcoin-community to have a comprehensive one number measure of security. It would help in discussions about blocksize and future discussions about network security, number of nodes, the whole Fee-market debate a LOT if one would have this chart.

Because besides talk about new cool services and the ever question of who satoshi is and the price of bitcoin the rest of the discussion are a very high percentage: network security, blocksize, number of nodes, mining profitability etc. And all those would be easiert to talk about if there was such a measure...

If there isn't such a site, i would still be interested in how the number is for todays network.

I guess it's more secured than we can imagine.