Earlier quoted context omitted.
Maybe this is where we diverge. > or the prosecutor looking to use the results of such an investigation as evidence in a trial I'm inclined to believe in the possibility of parallel constructions being used to cover up the best sources of intel. > Anything is possible.[...] As long as it's not cost-effective to trace the transfer, that's enough. It doesn't need to be mathematically impossible. And here, I think it is…
> I'm inclined to believe in the possibility of parallel constructions being used to cover up the best sources of intel. So am I, to a point, but even if they prefer not to disclose their actual methods (and are willing to commit perjury) they can't exactly hide the results. And others wouldn't have any incentive to keep their successes hidden. > ...I think it is probably cost-effective to come up with that technolog…
Yes, I think that is the fundamental problem with depending on 'mixers' against state-level actors. We both agreed earlier that the tech is theoretically possible. It seems like we're disagreeing about whether someone exists who is motivated enough to build the tech, and whether that person is also motivated to keep their tech under wraps.
> they can't exactly hide the results
It's a known method that US law enforcement has done in the past. Parallel construction is absolutely a thing. https://en.wikipedia.org/wiki/Parallel_construction
> the idea is to make all the transactions look the same so that even using your best graph theory tools you can't narrow down the possibilities enough to reasonably investigate all of them
I just don't get the impression that it's successful. There's a lot of 'metadata' that could be used to narrow the candidates down: geography, time, transaction amount, method of accessing the exchange (API / browser / desktop app ), age of wallets - I don't know which is specifically relevant here, but there's a lot of similar information which could be used to narrow the possibilities down, and most of it could probably involve 'fuzzy logic'. I just don't think that a threat model which includes state-level actors should ignore the possibility that transactions could be traced through mixers.
By the way, I'm really enjoying this discussion. Thanks for playing. :)