what i want to achieve is having a way select the ideal uxtos in some Stonewall TX
Great question! The bottom line is always going to be to wait as long as you can to get as many free remixes as possible.Let's say you coinjoin six outputs of 0.05 BTC, and then after a single mix you create a stonewall transaction which spends all six of your new outputs together. That's pretty bad for your privacy, and potentially allows all your outputs to be linked. This is going to be the same for any coinjoin implementation.
Let's say you coinjoin six outputs of 0.05 BTC, but then let each one get 5 free remixes before you stonewall them. That's significantly better and much harder to track.
I don't think you can pinpoint a number of mixes and say "After this number it is safe, before this number it is not." It's just not that simple. The anonymity set you get from each individual remix will differ. Some will have 5 inputs/outputs. Others can have up to 8 inputs/outputs. Each other output in one of your mixes might be spent immediately after that remix, which decreases your forward looking anonymity set. Conversely, every other output in one of your mixes might go through another 50 free remixes, which vastly increases your forward looking anonymity set. Similarly for a backward looking anonymity set - if all the other inputs in one of your mixes are very fresh, then your backward looking set is small. If the other inputs in one of your mixes have already remixed 50 times, then your backward looking set is much greater. Not only will the anonymity set change, but you need to consider your individual threat model as well. Are you wanting to hide your coins from your nosy neighbor next door? Or are you wanting to hide your coins from a nation state level? And how many outputs are your stonewalling together? Two might be OK after a couple of remixes; fifty UTXOs much less so.
You might be interested in looking at this: https://code.samourai.io/whirlpool/whirlpool_stats
As an example of my own practice:
I hit this milestone in Sparrow a few weeks ago, and decided to screenshot it for posterity:
I've added a few more to that count since then, and I've got a couple of such inputs in the same ballpark. It's turned in to an experiment now where I will just refuse to spend at least one of these outputs unless absolutely necessary just to see how high we can go.
As I mentioned above, just leave them in for as long as possible. Good luck to anyone trying to link those particular outputs together. I've added a few more to that count since then, and I've got a couple of such inputs in the same ballpark. It's turned in to an experiment now where I will just refuse to spend at least one of these outputs unless absolutely necessary just to see how high we can go.
