Meeting 2hr

meeting time: monero-project/meta #711

1. greetings

hello

Hello

hello

Hi

Hello

2. updates, what is everyone working on?

Busy making multisig txs to test #8149, with 3/5

So far looking good

Working on my understanding of BP

been working strictly on reviewing 7760 to understand its core logic. I finished stepping through old code while running the tests to understand exactly what the issues discovered are, and am working through understanding exactly how the fixes applied solve the tests

me: Working on seraphis enote scanning (thinking about how to handle reorgs today), plus did some miscellaneous robustness improvements to the seraphis library.

Released my analysis of the U.S. Federal Reserve data on use of cryptocurrency as a payment means:

Analyzing Dogecoin output spend age distribution over time. Some very preliminary result to share if we have time.

Afternoon, everyone :)

Hi all, im a developer but have never worked with Monero before. What's the best way to get started?

rustyrose[m]: is there something in particular you want to work on?

#monero-recruitment:monero.social lets not derail the meeting rustyrose

3. discussion, anything to discuss? comments/questions?

I guess that #8149 needs a final review, after the changes that happened.

Maybe we should make that known, to find some brave soul

The warnings about "experimental multisig" are merged and work wonderfully :)

Now only the code itself is needed ...

lol great

nothing in particular, but i dont know where to start

yes, please someone review 8149

UkoeHB: Sorry, I didnt have time to look at the multisig documentation but I promise I will get started this weekend.

rustyrose[m]: I'd start by learning something. Use software (maybe you'll find a bug or needed feature). Read github issues or open PRs. Read the code. Try to build something. We don't have managers or task lists, everyone is self-driven.

rustyrose[m]: ask in #monero-dev

dangerousfreedom: sweet that would be great :)

Speaking of needing more people to review critical cryptography, I am putting together a target list of cryptography-heavy universities and departments that the MAGIC Monero Fund could contact to fund Monero-specific research. Is anyone aware of a good list to start with?

I don't, but you could try contacting sarang for advice

Any other topics?

Like I've mentioned before, I'm looking at the blockchain of other transactional cryptocurrencies to see how stable the distribution of the age of spent outputs is over time. One way to do this is to fit parametric distributions to the 2020 data and then see how the fit performs when the 2020 fit is compared to the distribution in each week of 2021. This sort of simulates what a new decoy selection algorithm for Monero would

have to contend with: fit distributions on past data and then use it for future data.

So far I've looked at Dogecoin.

Here is a graph of very preliminary results:

L_FGT is a "privacy impoverishment" measure. L_Welfare is a "welfare" measure. Higher values indicate worse fit, i.e. higher difference between the "decoy" distribution and the real distribution during that week.

Each measure has two different "flavors". Basically, the higher parameter for each measure indicates that the fit was made with greater risk aversion, i.e. penalizes the distance between real and "decoy" to a greater extent as the difference increases.

Consistent with expectations, there are large spikes in the graph when there were spikes in the Dogecoin fiat exchange rate, number of on-chain txs, or amount of fiat-denominated value moved on the chain.

is it bi-modal? if you take out the connecting lines

Anyway, I'm not looking for any specific input at this point. Just letting you know the direction of this research.

I wonder if the age of inputs just moves with price...

ie, price increases causes older outs to move.

That'd suck for the fake out selection.

UkoeHB: Good question. I think it probably is. Basically, there is a "normal mode" that may reflect usual spend behaviors, and then when there is a speculative activity, the age distribution changes -- probably old coins are "waking up" and being sent to exchanges.

mitigating that kind of variance is big advantage to binning

moneromooo: It seems to me that price is a strong driver of shifts in the distribution, yes. I don't have the paper in front of me, but there was a paper a while ago that said that the majority of BTC on-chain activity was speculative, involving exchanges.

Actually, here is that paper: nber.org/papers/w29396

>We show that the vast majority of Bitcoin transactions between real entities are for trading and speculative purposes. Starting from 2015, 75% of real bitcoin volume has been linked to exchanges or exchange-like entities such as on-line wallets, OTC desks, and large institutional traders.

+1 this is building a stronger case for binning

Ok, maybe the paper isn't from a while ago. Oct. 2021 is the posting date

I'm also curious if outputs are younger or older than expected in these spikes

Why would one link "exchanges or exchange-like entities" to "on-line wallets" ?

I assume "on-line wallets" means something like mymonero.

Maybe because it moves the p-value down...

mixers?

Swapping services? They are not yet listed else

Oh. I am ignorant of what Bitcoin does so I used monero as a proxy. Alright then.

moneromooo: I'll see if they disaggregate the numbers further

jberman: Overall in my investigations of Dogecoin and Litecoin, it seems that the distribution of the age of spent outputs is less stable than I expected. That means that a static decoy selection algorithm that we have now (and will have with my near-term overhaul of it) would not do a great job of covering the distributional shifts over time. Which yes would probably improve the case for binning,

David Rosenthal reached similar conclusion blog.dshr.org/2022/02/ee380-talk.html

In the paper, "Figure 3: Decomposition of real volume" shows that 40% of BTC volume involved exchanges, narrowly defined (I think)

only 27k "economically meaningful" bitcoin txns/day

75% are just inter-exchange\

only 2.5% of bitcoin txns are p2p

hm, he's just referencing the paper you linked

One caveat here: In theory Monero's adjustments of the decoy selection algorithm based on number of txs in each block would mitigate this issue to some extent. I did not try to mimic that effect and adjustment in my work with Dogecoin or Litecoin so far.

It was clever to implement that adjustment :)

This is true, it biases towards greater volume

Just curious, is there something that shows the spikes are caused by a higher freq of older outputs being spent and I'm not seeing it?

jberman: No. These graphs are a high-level summary. I will work on quantifying exactly how the distribution shifted. Mean, median, variance, skew, and probably a visualization.

Cool :) I could definitely see spikes caused by a higher freq of younger outputs too. Rabid frenzy to buy/sell/move between exchanges/off to wallets/test txs etc.

One way to think about these lines in the chart is that they show a metric that is akin to the Kolmogorov-Smirnov distance between the 2020 distribution and the distribution in each week of 2021. So it is a measure of difference, but it doesn't show in which ways, exactly, it is different.

I haven't really commented as I haven't had much to say, as my work is mostly around Monero not on it, though I did want to note the burning bug has had been a topic of mine recently, before this meeting ends. I have an open MRL issue documenting an adjustment to the shared key definition removing it as an option, along with a suggested alternative by koe, which isn't planned to be adopted (and after further review doesn't

appear possible due to DoS issues). Seraphis itself has the update, and I've reached out to a few parties to discuss the burning bug, the evolved attack, and mitigations to ensure not only are we forward thinking, yet making sure no one missed the original memo.

Sorry for interrupting the current conversation with that :p

no worries, thanks for commenting

ok that's the end of the meeting, thanks for attending everyone

Only the drama was missing :)

shaping up to be a calm, productive day

UkoeHB: Don't jinx it :o

jberman: Here the mean, median, standard deviation, and skewness of the Dogecoin 2021 spent output age distribution:

Spikes in the mean line up pretty well with sudden worsening of the "static" fit from 2020:

So the hypothesis that old outputs "wake up" to take part in the speculative activity on exchanges is supported by the data, I think.