Quote from: o_e_l_e_o on April 30, 2022, 07:29:06 AM

A way of doing this, which is trivially easy when the coordinator is working hand-in-hand with a blockchain analysis company as is happening here, is for the blockchain analysis company to put forward a few dozen of their own inputs to be coinjoined. The coordinator then picks (for example) 19 inputs belonging to the blockchain analysis company and 1 input belonging to you. You think your outputs are obfuscated, but the blockchain analysis company knows exactly which ones are yours by process of elimination.

What you are describing is called a targeted sybil attack, which is a known attack vector of all centrally coordinated coinjoins. In fact this is the only sybil attack that as far as I know is feasible to do, although it'd still be noticeable and it's still not a cheap attack. The coordinator has to provide hundreds of inputs and outputs (equivalent to hundreds of normal bitcoin transactions) to deanonymize a single UTXO. In fact double that cost as the attacker must prepare the UTXOs in advance by making transactions. Furthermore, it's also detectable by looking at the remixes in the coinjoin. If the remix ratio is different from other coinjoins then we would know the attack happened. You might think then the attacker should spend more money on fees to pre-mix coins to make the remix ratio perfect. But even then there would be a pattern with the timing of the pre-mixes that would clearly differ from normal coinjoin volume. Executing such attack in an unnoticeable fashion would cost an extraordinary amount of resources since to hide all the patterns the attack would have to keep participating in every single coinjoin with many inputs until the target UTXO would decide to mix so the attack can be cleanly executed.

This has nothing to do with blockchain analysis tho.