As much as I like this and want it to succeed, the amount of protection darksend gives you is currently rather pathetic. Take a look at this (
http://explorer.darkcoin.io/tx/b9d3850d62e0a6c460e220713710fda602d5daea4771af1bbb397b35e3b6eb5f ), or any other darksend transaction, and you'll see it's very easy to pick apart. The first input clearly goes to the first two outputs, as they're the only ones less than the input, and the second output can be seen to be the change address, as it has too high precision for darksend. The second input goes to the fourth and sixth outputs, as they add up to the input - 0.01, and no other combination adds up to an appropriate amount, and yet again, which address is the change address can be seen by the precision. This leaves the third input to the third and fifth outputs, which add up to 0.1 less than their input.
Eduffield's idea of pooling only same amounts is excellent and would fix this, but in the current state this just isn't effective at all.
It's actually not as pathetic as you think it is. It's only the foundation (with structure installed for the other aspects, that are not turned on yet). We're just trying to stress test the foundation of what is to become a full darksend implementation. People are getting a bit too vocal in their excitement and though it is indeed a huge step forward, it's far from complete.
I hope my suggestion about converting all payments into combinations of common amounts will be used. Then it becomes much harder to establish connections. Also maybe limit it to sending round numbers without fractions. Not sure if adding some random delay would also help, it would slow down the transaction. Another idea is to have the smaller amounts aggregated into the next larger denomination and forwarded onto the next block. By the time the large denominations is mixed, it will be composed of many smaller transactions from a variable amount of time in the past.
Keep in mind when there are a small number of people participating, it is trivial to solve the puzzle of who sent what to whom. Now imagine thousands of transactions being split into more transactions being partially combined and forwarded onward.
Of course, nothing will work against big agencies (they can probably directly tap into your computer anyway) but to keep random people from using blockchain to figure things out, this is definitely achievable using this type of method
James
