Quote from: n0nce on January 28, 2024, 10:24:54 PM

Couldn't the coordinator just use own inputs and 'prioritize' them over real user inputs whenever they need to do a sybil attack? How would that be obvious to other users? It could easily be that those are not coordinator inputs but that it's real user demand that's simply higher than usual for a brief period of time, no?

if the number of inputs exceeds the maximum number set by the coordinator which I think is 400 for their new protocol, the coordinator would automatically arrange another round.

But then we really need to reach some ground on defining the purpose of sybil attacks, be it those done by outside attackers or the coordinator itself, the point of sybil attacks in a coinjoin is to bring the anonymity set for the victim to 1, it's the only possible way to link x input to y output -- otherwise, the attack would only reduce the anonymity set/score.

The coordinator could certainly force you into a conjoin that has 399 inputs it owns and 1 is yours, but that means, there will be a coinjoin for every "real" participant, while that is doable in theory, it would certainly raise the flag (you can register 2 different inputs and see if they end up in different rounds every time you do that).

Also, round status is publically available through Wasabi API, you can acquire the current input count, if the current round is at the input registering phase and has 50 registered inputs, and then you try to register 2 different valid inputs and they end up in a different round, you know they are doing something fishy.

Obviously, if they have a targeted input (or a few of them) to which they want to link -- then that would make sense, and they could be doing that already and manage to hide it but a full-scale Sybil attack is just not feasible IMO.

Quote

I don't think users get something like a timestamped proof that they submitted inputs to a CoinJoin at a certain point in time (could be used to show that they entered the CoinJoin before the coordinator started to attack), right?

timestamped proof? not sure, A proof, yes, the coordinator creates Tor identity at input registration, obviously the person who receives it knows the time at which they received the credential, I don't think they can prove it to someone else.

With that said, I am not claiming Wasabi don't/won't do any of that, I am just stating that it would be very difficult to hide a sybil attack, besides, Wasabi doesn't need any more criticism, you make chain analysis scums richer every time you use it.