Avoid selecting coinbase outputs as decoys about research-lab

[Rucknium]

@SChernykh has announced an upgrade to P2Pool that will reduce the number of P2Pool payout outputs by 50 to 67 percent. That's a big accomplishment, but user privacy can still be improved by this proposal to avoid selecting coinbase outputs as decoys.

I did some historical analysis of P2Pool payout outputs and how they affect privacy of non-mining users.

Below is a plot of the share of blocks mined by P2Pool. It has risen slowly over time to about 9 percent today, with some spikes and dips.

The total number of Monero transaction outputs per week fell after the hard fork because of the crash in Monero transaction volume.

The post-fork crash after the hard fork plus the stable P2Pool share of mined blocks caused the share of P2Pool outputs to increase from about 5 percent to 15 percent.

The low share of P2Pool outputs before the hard fork is good news. It means that the negative privacy effect was low at a time when the ring size was lower. The pre-fork ring size was 11. Now it is 16.

We can approximate the effective ring size for each week using the theoretical Binomial distribution again. Effective ring size is the number of ring members that are not P2Pool outputs. The median effective ring size is plotted below. The median value of X means that 50 percent of all rings have an effective ring size of X or below.

The plot shows that median effective ring size still increased after the hard fork despite the larger P2Pool share of outputs.

The median is "typical luck". We also want to be aware of what can happen when an unusually high number of P2Pool outputs are selected as decoys. These are "unlucky" rings. Below is a plot of the 5th percentile of effective ring size. That's the 5 percent unluckiest rings. The unluckiest rings have almost the same effective ring size before and after the hard fork.

The code to reproduce this analysis is at https://github.com/Rucknium/misc-research/tree/main/Monero-p2pool-Output-Stats/analysis

[WojasKrk]

W #78 #78