Evaluating the privacy of PrivateSend

Dashcoiner said:
This FUD about PrivateSend has been discussed and clarified many times. We don't have money to waste on another study proving what we already know.
Click to expand...
Hey Dashcoiner.

Can you provide some links to the other studies that you are talking about? In my opinion even if no transactions can't be de-anonymized it doesn't mean that PrivateSend is flawless and there is no room for improvements.

Have you seen any other analysis that can give you the most probable mixing sources even when a transaction can't be de-anonymized?

Did you check my analysis of the de-anonymization contest transaction?
 
Last edited:
akhavr said:
Can you explain why you named the percentage axis a Traceability? I'm afraid, I don't get it :(
Click to expand...
Thanks for asking and no need to be afraid because I'm not very good at explaining stuff..

For example, for this transaction the value would be 0.5394215404817816 for 2 rounds, 0.25030889621087315 for 3 rounds etc. So when there is multiple transactions analyzed in the charts, the average of those values was used. For example that transaction has two 0.01 inputs so it would be included in the blue lines at the x-point of 2 inputs. In the column charts it would be included in the blue 1-500 height difference columns because the two inputs (this and this) have a height difference of 8 blocks.
 
Perhaps I'm not caffeinated enough this morning yet, but I still don't get.

Can you define traceability maybe? To my imagination it is a value, correlated to the probability of identification of a (set) of source txids for any defined output, after mixing. (As you might see, I'm also not good at explaining ;) )
 
akhavr said:
Perhaps I'm not caffeinated enough this morning yet, but I still don't get.

Can you define traceability maybe? To my imagination it is a value, correlated to the probability of identification of a (set) of source txids for any defined output, after mixing. (As you might see, I'm also not good at explaining ;) )
Click to expand...
Among the set of all possible mixing sources (create denominations transactions) of a PrivateSend transaction it is the probability of the most probable mixing source (create denominations transactions). The probabilities are calculated by repeatedly performing random walks from the PrivateSend transaction in the PrivateSend graph with the limitation that transaction outputs can only be traveled once in multi-input PrivateSend transactions. And the PrivateSend graph is a DAG where the root node is the PrivateSend transaction, leaf nodes are create denominations transactions and the intermediate nodes are mixing transactions.

So for example, when a PrivateSend transaction has a "traceability" of 0.5 for a given number of rounds, it means that it's possible guess one of the mixing sources and with 50% probability the guess would be correct if the assumed number of rounds is correct. In reality it is almost never possible the know the number of rounds with a few rare exceptions.

Maybe there is a better word to describe what I mean. And by the way I should have used "Average traceability" in the charts because they are all averages.
 
Antti Kaikkonen said:
So for example, when a PrivateSend transaction has a "traceability" of 0.5 for a given number of rounds, it means that it's possible guess one of the mixing sources and with 50% probability the guess would be correct if the assumed number of rounds is correct.
Click to expand...

Does that mean, that 1 in graphs above mean 100% probability of the correct guess, given the assumed number of rounds is correct?
 
akhavr said:
Does that mean, that 1 in graphs above mean 100% probability of the correct guess, given the assumed number of rounds is correct?
Click to expand...
Yes. It means that at least one of the mixing sources appeared in all of the randomly simulated mixings. But of course it is easily possible that a different number of rounds was actually used. It is also possible that different inputs were mixing using a different number of rounds.
 
You must log in or register to reply here.
Back
Top