Hacker News Reader

Comment by ontoillogical | Hacker News Reader

> And in any event, bitcoin is never going to work as a global payments system. Not only does it suffer from having a slow-growing money supply and a metastasizing transactions file which has to live on every user’s computer, it also encourages destructive computer hacking. The way that the money supply grows, in the bitcoin system, is by people harnessing the power of hundreds or thousands of computers to solve very complicated mathematical tasks, earning bitcoins for doing so along the way. And the easiest and cheapest way of doing that is to do so illegally, by stealth: set up a “botnet” of hacked computers to do your bidding for you. The incentives, here, are very bad indeed.

I'm not an expert, but I thought that these days you basically need specialized hardware/GPUs and a traditional botnet isn't going to get you that far. Am I right?

View on HN

You are right. CPUs are already useless for mining. Currently, GPUs can still mine their share (but running a GPU miner on a botnet is a risk; computers become very noisy, hot and slow, which increases the detection rate). But as the first ASICs have started shipping (or have they?) this is expected to change very soon, and the only way to make a profit with mining will be with custom hardware. This rules out botnets pretty effectively.

If botnets are free to operate, then won't they still be used? They're only less efficient than ASICs if you have to pay for electricity, right?

Botnets are usually rented out. So there is an opportunity cost, when the nodes and their computational/network resources could be used for more profitable activities.

I did a back of the napkin calculation the other day to see if a botnet could reasonably reach the 50% computing power required to overwrite transactions. The largest known botnet (BredoLab @ 30,000,000 bots) with modest hardware could have potentially done it. This computational power is a multiple of the combined computing power of the top 100 supercomputers in existence today.

Of course, botnets are frequently on older machines, so the actual computational power of BredoLab probably wouldn't have been enough. Also, as popularity increases, this will become less reasonable.

And also, anticipated by Satoshi, was that the long tail of smaller botnets that would be unable to double spend would take to bitcoin mining because it would be more profitable, so increasing security and making it much harder for a single large botnet to take over.

There eventually wont be much point to mining when the cap is reached, then there will be a point to process blocks for their transaction fees. You are also right about the hardware issue, plus there are other more profitable ways to use a botnet. There is also just a general problem with the argument that it incentivises botnets, they are their own incentive, anything that needs bandwidth, cpu power, and stolen information will be much easier with a botnet.

> I'm not an expert, but I thought that these days you basically need specialized hardware/GPUs and a traditional botnet isn't going to get you that far. Am I right?

How many computers within that botnet will have the appropriate GPU? That's not counting the computing power of the network as a whole either; Imagine something like BitcoinMiner@Botnet. I think it could easily be lucrative enough to make it worthwhile.

The largest known bot net had ~30,000,000 computers at that scale CPU's work just fine. Not that it's the only thing to do with a bot-net of that size, but it's clearly a lower-bound where if your not doing a denial of service attack or sending spam you might as well mine bit-coins.

Assuming the average CPU runs around 2 Mhash/s that's 60,000,000 Mhash/s which is the equivalent of about 75,000 ModMiner Quad @800Mhash/s which would cost 75 million, so you would probably still top the mining charts.

Modest CPUs produce bitcoins very slowly. So, there is an incentive to create a botnet to harness the power of thousands of CPUs to mine bitcoins.

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

There is no point in even putting together a botnet with thousands of CPUs anymore, a couple of FPGAs or an ASIC based solution would run rings around that botnet. Ok, an ASIC will cost a bit of money and a botnet is free but the difference in performance is such that only a very large botnet would be worth it and those can probably be monetized more effectively in other ways.

Botnets make a ton more money by just ad-clicks anyway. I bet its more beneficial to fraudulently ad-click with botnets than it is to mine bitcoins.

Reply on news.ycombinator.com