Quote from: o_e_l_e_o on May 01, 2022, 05:50:07 PM

I can produce a hundred outputs from a hundred separate transactions for less than $10 at current prices and fee rates.

That might be a bit too low, but perhaps. There's no disagreement here. If you read what I wrote, I made the same point: "in fact this is the only sybil attack that as far as I know is feasible to do"

Quote from: o_e_l_e_o on May 01, 2022, 05:50:07 PM

Add in a variety of sources, intermediate transactions, mixes, coinjoins, etc., along the way to obfuscate things, and you could still do this for a trivial amount of money. And in an ongoing attack, outputs from previous rounds can be recycled in to inputs in new rounds.

There are many patterns you must copy to execute this attack unnoticeably. For example, let's assume a pattern that 10% of all inputs of a coinjoin are coming from the preceding coinjoin. In that case, to copy this pattern, you would have to be 10% of every single coinjoin, which is already far from "a trivial amount of money" and this is just a single pattern example.

Quote from: o_e_l_e_o on May 01, 2022, 05:50:07 PM

The coinjoin fees would be irrelevant since you can just hand them straight back to the blockchain analysis company since you are paying them for their services anyway.

Although the main point you're making that coinjoin fees aren't relevant, is correct, you box it into a loaded statement through a second incorrect claim, which is that "all the money is going to a blockchain analysis company." Loaded statements is quite a disturbing pattern I'm noticing in our conversations, so I'd greatly appreciate if you'd stop doing that.

Quote from: o_e_l_e_o on May 01, 2022, 05:50:07 PM

Beyond the scope of the average user, and we can't trust you to tell your users since you are currently selling them out to a blockchain analysis company so you can serve institutional clients for your own profit.

We're not selling anyone out. We wouldn't even have the ability to do so even if we wanted to. Another loaded statement aside, it'd indeed be a problem if the detectability of an attack would rely on trusting the attacker itself. Cheesy

Luckily it's not the case.