The first argument everyone makes is that miners would switch pools if any pool started censoring transactions. This may be true, but to keep an objective measure of security we must assume they would only switch if it were profitable to do so.
With that objective measure of security, what about this attack?
1) Create n new pools
2) Provide a 0.1% subsidy for blocks found by each of the new pools
3) By definition, all miners would switch to one of your new pools, because they would maximize profit by doing so
4) ? ? ?
5) Profit
The attacker would have 100% of the hash rate, for a cost of $8640/day. It sounds absurd because it is. Or is it?
Also, what about this bounty to Bitcoin XT blocks, currently 1.3 BTC? https://cryptoplay.net/vote/ Why hasn't it been snapped up? The "objective measure" rule would say this free money should not remain on the table.
Or this bounty? https://www.bigblockbounty.com/ Would this line of thinking predict that the 4+ BTC reward for 100 XT blocks would be quickly claimed?
There are several things going on here:
1. The rewards offered are driven by donations and not a long-standing and sustainable offer
2. The cost to update the mining pool software is greater than the one time reward.
If those bounties were a promise of $8700/day divided among conforming blocks and you have a mining pool that finds 1 block per day on average, then you will double your money by conforming. Except that it is not just doubling the money earned by this miner (assuming 5% margins) this particular bounty would be increasing the profits of the first miner to switch from $300 to $9000 which represents a 30x increase in profitability by switching.
So clearly someone will take that bate and produce 0.7% of the blocks. The market never likes to leave anyone with such high margins, so miners will switch from the main network to the conforming network (merging mining) until the profit from mining under the alternative rules falls to some threshold. If you assume miners will continue switching so long as it doubles their margin from $300/block to $600/block then an attacker will have gain 20% of the miners for the cost of 0.7% of the hash power.
The market will then provide feedback to the attacker which can increase their daily pay until all miners switch.
Providing incentives to attack only makes sense to the extent that it is cheaper than than undermining the support of the main chain. Stated another way, offering to double your salary may not get as many defectors as threatening to make your salary negative. So long as you are earning a safe income and able to pay your bills you can entirely ignore the attacker.
Enter negative mining. For $8000 per day spent on negative mining you can reduce the income of everyone else mining in public pools by 50%. (For $5000 per day you can cut income by 20%) Such a huge decline in income would cause almost all other miners to be operating at a loss. Once this happens these miners turn off their equipment. They have no financial incentive to SPEND money to defend the network.
So if your ONLY goal was to stop block production, then it would only cost you between $5000 and $8000 per day.
Now if you have a lot of mining equipment and no longer had the ability to mine in the pools and the attacker offers you your old job back *IF* you mine for them now it is a question of 0 income and a complete loss on your capital investment or some income. It now becomes clear that not joining the attacker is irrational from a profit-seeking point of view. You could walk away in protest (not wanting to take dirty money) but that only harms you and doesn't help the network.
Now this is where it gets really interesting... an attacker can approach the largest mining pool and make them an offer they cannot refuse, "You can either censor transactions and make more money, or suffer a negative mining attack that will put you out of business". Once a single pool caves into the demands, the that pool can be used against the other pools. If no pools cave, then the attacker needs to rent hashing power for his attack.
Conclusion: The Bitcoin network is only as secure as the profit margins on mining, not the cost of mining. The lower the profit margins fall the cheaper it becomes to perform the negative mining attack.
Mining pools would have to come up with a defense against negative mining which is not trivial.
Topic: Bitcoin 100x less secure than commonly believed (Read 8449 times)
