Logs for the Monero Research Lab Meeting Held on 2019-03-11
Network upgrade, Point release discussions, MRL work, and miscellaneous
dev diaries
crypto
research
el00ruobuob / sarang
Logs
<sarang> Our meeting begins presently <sarang> Let's go ahead and get started. Agenda is here: https://github.com/monero-project/meta/issues/314 <suraeNoether> howdy everyone <sarang> 1. GREETINGS <sarang> hi <MRL-discord> <Isthmus> Hello! Biking, in soon. <parasew[m]> hello! <sarang> Let's recap 2. NETWORK UPGRADE <sarang> Kudos to everyone for a successful first upgrade <sarang> I don't recall when the second was slated to occur, since block arrival was stunted <sarang> Any thoughts on the upgrade after the fact? <xmrmatterbridge> <rehrar> Hi <sarang> I believe it was dEBRUYNE who wanted an upcoming meeting specifically to talk more deeply about the future of PoW <parasew[m]> anyone monitored the "old chain"? if there have been this large amount of asics on there, and not turned off it should be visible <sarang> I believe sgp_ ran some blackball numbers on it <sarang> and found essentially nothing of interest <sarang> but as far as hashrate, I am not sure <sgp_> yeah, no chain reactions so far, very few known spent outputs through reused key images <sgp_> impact on network privacy so far is essentially 0 <sarang> sgp_: were the key image reuse numbers for only v9 and v10? <sgp_> yes, just those two <sarang> great, thanks <sarang> Relating to this, we can also introduce 3. NEXT POINT RELEASE <sarang> Not all desired non-consensus changes made it in to this release, so Sometime Soon (tm) will be a point release <sarang> BP optimizations will be one nice addition <sarang> I would like output selection to also be included... we talked about it at length at an earlier meeting <dEBRUYNE> sarang: Correct. It's a topic with a lot of depth that requires an extensive discussion imo <sarang> suraeNoether: do you have a current recommendation for output selection? <suraeNoether> i'm running into problems testing the matching code, based on this problem too <sarang> Here is a discussion of the different algorithms: https://github.com/monero-project/meta/issues/307#issuecomment-466514757 <suraeNoether> iirc the output lineup method performs quite well <sarang> I prefer it among the others that were tested <sarang> But it's a change that deserves more than two thumbs-up :) <suraeNoether> there is no optimal solution, but some solutions are better than others and the output lineup method is more reasonable than the other proposals, and i have no new proposals to make (yet) <sarang> I updated the sim code (link in agenda) to examine the output weighting in more details <sarang> Hopefully the BP optimizations are less contenious <suraeNoether> uhm i think i have one possible proposal that i want to chat about with you by side channel to hash out some details <sarang> sure <sarang> We should have a formal recommendation before whatever date is set for the point release code freeze <sarang> Anything else relating to the point upgrade that ought to be discussed? <sarang> ping moneromooo perhaps <xmrmatterbridge> <rehrar> I just want timelines. Nothing to say on content. <moneromooo> hi <moneromooo> What's the question ? :) <moneromooo> I don't know about any date. Depends when we get all the stuff on master ready really. <sarang> Anything relating to the next point release you'd like us to discuss? <moneromooo> None that come to mind right now. <sarang> ty <sarang> In that case, let's move to 4. ROUNDTABLE <sarang> suraeNoether: care to go first? <sarang> OK, I can go first instead <suraeNoether> ok <sarang> aha, go ehead <suraeNoether> heh <suraeNoether> Well, my simulations for the matching code are to the point where i'm running a matching on some test data now to generate a confusion matrix. <suraeNoether> i'm also editing the manuscript describing the whole process <suraeNoether> one of the problems i'm running into is actually simulating our output selection in part because it's not clear which direction we are going yet <suraeNoether> and it occurred to me that this could help inform our choice of output selection by seeing if one of these possibilities makes matching easier or harder <sarang> IMO matching expect spend with proper weighting seems optimal enough from a purely timing perspective <sarang> (leaving out questions of binning etc) <suraeNoether> when i say easy or hard i don't mean in terms of time, because as we've seen matching is essentially super duper fast <suraeNoether> i mean in terms of false negative and false positive rates <suraeNoether> but you are 100% on that <sarang> aw shucks <suraeNoether> i'm working on a variety of other side things but i'm shooting for this matching paper to be complete and published some time in the next 2 months <sarang> Excellent <suraeNoether> if we get more speakers for the konferenco, then i won't be speaking, but otherwise i will probably be presenting on this at the konferenco <sarang> Neat; anything else of interest to share? <suraeNoether> that's all i have today, thanks! <sarang> Righto <sarang> I have a few things <xmrmatterbridge> <learninandlurkin> The line up is looking great btw! Fantastic effort for a first konferenco <suraeNoether> catching up on lots of reaidng in algebraic geometry :D <sarang> First, my next FFS/CCS will be posted soon <sarang> As was discussed here, in -community, and elsewhere, the request will be for immediate payout <sarang> This means both donors and I know the actual value of the donations <sarang> Since this is a big change, any questions or comments on it? <sarang> (presumably suraeNoether will be doing the same arrangement) <suraeNoether> i'm in support of this, and i will indeed be mimicking this <sarang> Folks who do not trust us to run with the money should, of course, not donate <sarang> But my hope is that our records have shown we're good for it :D <binaryFate> happy we came to that solution eventually, hopefully will be better for your guys <sarang> Thanks to binaryFate and others for agreeing to this change <binaryFate> yes the idea is that donors being careful should discourage randomers to do the same <sarang> The CCS posting will _very_ clearly state the arrangement, so there is no confusion <binaryFate> If you figure out the markdown <sarang> Yes indeed <moneromooo> Technically, it's within the existing rules as stated: one milestone, which consists of "sarang starts working" :) <sarang> Second, the paper that suraeNoether and I have been collaborating with external researchers on (DLSAG et al.) is in final review now <sarang> We've been asked not to share it before it's released as a preprint, as a courtesy to all authors <suraeNoether> *nod* <sarang> It has some great details on useful constructions that I'm sure we'll discuss at length after the preprint goes to IACR <sarang> it'll be submitted for a conference as well <sarang> Third, I wrote up some additional tests and code for Bulletproofs MPC <dEBRUYNE> sarang: How does this work if the proposal is not fully funded yet when your period starts? <sarang> Two options: either the bulk is paid out and it stays open until filled <sarang> or it all sits there until fully funded <sarang> I prefer the first, but am open to discussion <sarang> Regarding Bulletproofs MPC, real_or_random had some great thoughts on this before the meeting (but I won't put him on the spot) <suraeNoether> i imagine that the important part is laying out which way it goes in the proposal <sarang> the question has to do with what a malicious player can do <sarang> We chatted about the fact that an evil player could try to pull what amounts to a cancellation of partial proof elements, effectively setting the inputs to the hash that generates a F-S challenge <sarang> I couldn't find a way that this could be used as an exploit, aside from obviously generated an invalid proof <sarang> but the security proofs for BPs do require that F-S challenges are uniform <sarang> I had neglected that point when I had thought about this earlier <sarang> My strong suspicion is that proof elements are still uniformly distributed in the presence of a dishonest challenge due to the prover's randomness, and that you still get zk in this case (but not provably) <sarang> Moral: if we do anything in the future that requires/desires this scheme, these things would need to be considered <sarang> Any questions/comments relating to this? <sarang> allrightythen <suraeNoether> i think we should continue to ponder it and write something up formally about the BP MPC schemes <sarang> Well that's the thing... there's really nothing to write formally <sarang> You can probably solve all the theoretical woes by having all players commit to their proof elements before multicasting them <sarang> then an honest prover is guaranteed uniform F-S challenges <xmrmatterbridge> <learninandlurkin> Sorry but I'm a little out of the loop here. What exactly are BP MPC for? something to do with multisig with BP? <suraeNoether> it's nice to think about collectively computing BP range proofs, but I'm still v curious about the coinjoin approach that we are considering on the larger scale. <sarang> Ideally, untrusted parties could generate single BPs for outputs <suraeNoether> after all, it's hard to even think about threat models unless we know how these things will be used in practice <sarang> Sure, this is all pie-in-the-sky right now <suraeNoether> learninandlurkin: collaborating with friends to compute a range proof for a coinjoin style transaction, so that the participants don't reveal their amounts to each other <sarang> But yes, the threat model would be very different depending on how the rounds go <sarang> Finally, suraeNoether had shown me this a while back: https://lelantus.io/lelantus.pdf <suraeNoether> agreed on the commit-and-reveal; expensive but usually does the trick to ensure participants can't be rewound inappropriately <sarang> An interesting application of some of the fundamentals behind Bulletproofs and the old StringCT scheme <xmrmatterbridge> <learninandlurkin> So... allowing multi-input transactions where each user doesn't know the amounts of the other inputs? Sounds useful <suraeNoether> learninandlurkin hence our interest in nailing down threat models *nod* <sarang> I've been playing around with some of the math in that paper to see what nuggets could be extracted <suraeNoether> oh i had a brief thing to point out: isthmus and n3ptune at noncesense-research-lab answered one of my requests and we now have a complete empirical distribution of number of inputs and outputs per transaction <suraeNoether> forgot to mention this: <sarang> Neato, where is this distribution to be found? <suraeNoether>https://github.com/noncesense-research-lab/tx_in_out_distribution <suraeNoether> the data surprised me <dEBRUYNE> <sarang> I prefer the first, but am open to discussion <= I'd be OK with the first, but perhaps it would be most convenient to use a rounded number <dEBRUYNE> e.g. if 211 XMR is funded, pay out 200 <sarang> You won't believe what's in tx_distribution_in.csv! <dEBRUYNE> Mebbe malware <dEBRUYNE> :P <suraeNoether> super heavy tails for one thing, and a rootkit for another <sarang> dEBRUYNE: perhaps a full payout at date X, and then a second payout at either date Y or completion, whichever comes first <binaryFate> <sarang> I prefer the first, but am open to discussion <-- donors will have no incentive to fund in time, it will drag till the end of the period <sarang> binaryFate: how would you do it? <binaryFate> I like the incentive to donors of you proposing something and getting to work on it only if funded <xmrmatterbridge> <learninandlurkin> I imagine coinjoining going on would really complicate output selection. Or is there some idea where they work off each other to get rid of heuristics? <sarang> Depends on how timely it is <suraeNoether> learningandlurkin coinjoin brings a whole new nightmare to the party. does everyone bring their own mix-ins? certainly nothing is to stop a malicious party from coinjoining with a bunch of badly selected mix-ins <moneromooo> A ring is one person only. Fake output selection is untouched. <sarang> Well each input signs with its own ring <sarang> ^ <moneromooo> That person makes their own ring, yes. Otherwise others would know which is the real out. <sarang> The benefit is breaking the assumption of one-party control of outputs and the link to the input rings <binaryFate> What about simple attack of using the same 10 decoys as one of the other participants? <suraeNoether> ^ <msvb-mob> Is parasew, nevvton, or txmr in the channel? <binaryFate> mmm you don't know which are decoys, nevermind ^^ <sarang> If this moves forward, hopefully we can determine the necessary practical security for BPs <sarang> If we can't aggregate, they'd have to be separate for each output <suraeNoether> my beard is getting very thoroughly stroked this morning. much to think about... <sarang> I believe we'd get practical security without player commitments, but not provable <sarang> Anyway: does anyone else wish to share interesting research before we close? <xmrmatterbridge> <learninandlurkin> Yes it sounds like the interplay between coinjoin and ringsigs will require some diagrams for me to ever understand. Could get complicated. <suraeNoether> i think you would want a commit-and-reveal stage for everyone to see the ring members to prevent malicious ring intersection in the coinjoin <sarang> MoneroCoinJoin: an easy 14-round process! <suraeNoether> isthmus and i have been chatting about methods of extracting the true spend-time distribution from the monero blockchain without knowing exactly which outputs have been spent <suraeNoether> that's a very nascent conversation, though I think it'll end up being a very straightforward project <sarang> Discussions in #noncesense-research-lab I presume? <xmrmatterbridge> <learninandlurkin> so, truish spend-time distribution <binaryFate> Are there regular meetings on this or just continuous discussion? I had been working on this at some point and have some code around aiming to graphically show the real spend distribution <sarang> I've seen a few informal conversations in #noncesense-research-lab but didn't know if suraeNoether had something more formal <suraeNoether> binaryFate: ah, no, this has been a casual conversation by side channel, but there is clearly interest <suraeNoether> i'll start blabbing about it in here more publicly <sarang> In the interest of time, let's review 6. ACTION ITEMS and then close to continue discussion afterword <binaryFate> Ok don't hesitate to ping me on this <sarang> I will be posting my CCS request soon, tidying up the output selection stuff for a recommendation, getting the DLSAG application paper reviewed and out the door, and playing around with that Lelantus paper when/if I get a chance <sarang> suraeNoether: ? <suraeNoether> CCS request, working on simulations and measurable numbers for matching, and looking into using our matching code to answer questions about output selection <sarang> excellent <suraeNoether> also casual github maintenance <sarang> Any final questions or remarks before we adjourn? <xmrmatterbridge> <learninandlurkin> once you guys have made a recommendation for output selection <xmrmatterbridge> <learninandlurkin> and it gets implemented, what's the next big focus? <sarang> There will be much to consider in the realm of refund and payment channels <xmrmatterbridge> <learninandlurkin> Ooh yes the refund ideas from a while back were really interesting <sarang> and some aspects of output selection, like linking spends across rings in txns, is not solved yet <xmrmatterbridge> <learninandlurkin> Seems like a logical next area of research <sarang> and if coinjoin works out, there will be a lot to consider with that <sarang> Also transaction relay and network-level anonymity stuff that's still in progress <sarang> To quote the Simpsons: "like the cleaning of a house... IT NEVER ENDS" <sarang> But on that note, our meeting does end <sarang> Thanks to everyone for attending. We're adjourned; let the conversations continue