14:54:08 Goodell, Salazar, & Slaughter (2024). “Uniformly Most Powerful Tests for Ad Hoc Transactions in Monero.” has been posted: https://github.com/cypherstack/churn/releases/tag/final 14:54:41 Here is my review of it: https://github.com/cypherstack/churn/issues/2 18:39:07 Rucknium: do you have any actionable takeaways from this paper? I have read it twice and I've read your notes, and I haven't found anything actionable yet 19:31:27 shall we list the paper on getmonero.org/resources/research-lab/ ? 19:39:55 cc Diego Salazar Rucknium 19:40:36 We're writing some stuff up for a blog post about it. 19:43:01 Right now there's a general understanding that the more churn there is, the better for everyone. But obviously this needs to be weighed against blockchain bloat. 19:43:49 so we shouldn't list? 19:44:12 Oh, go ahead. Though we may do some small updates. Will ping you if/when that happens. 19:44:20 Note the Salazar listed there isn't me ;) 19:46:08 Salazar the Monero family? 19:46:42 My brother has a master's in math and has looked at cryptography and Monero off and on due to my own interest. 19:47:05 My other brother is a good dev and has done stuff with Stack Wallet in the past also :P 19:47:51 so not too far off 19:48:06 *le mafia* 20:13:36 <3​21bob321:monero.social> Don diego 20:19:29 Where will it go up on? https://blog.cypherstack.com/ https://cypherstack.com/blog don't work. :P 20:19:46 y eah I'll make a blog for the site. 20:29:28 From what I understood, this article expects churns to blend with the standard decoy selection perfectly. But real churns will be done manually in 99% of cases, and they will not follow the perfect decoy distribution. 20:57:26 sgp_: Not really. It reaffirms that having a decoy distribution be the same as the real spend age distribution is best for privacy, in this churning paradigm too. But of course I would say that. This paper says that even if you have ideal churning (which we aren't sure anyone actually does), there is a way to try guess that a tx is a churn with a high rate of error. 20:58:39 I have compared EAE/churning research to the hunt for Moby Dick. IMHO, the white whale is still swimming freely in the deep blue sea. 20:59:41 I definitely didn't think we'd "solve" this churning thing when no one else could. Just wanted to keep building. FCMP is Soon (tm), but it could still be a ways off. 20:59:48 Was looking to see what options users had in the meantime 21:00:28 Getting actionable data on churning (that is comprehensive rather than situational) seems even further off than FCMPs, haha 21:00:45 IMHO, this is a component of a multi-component attack that an adversary would use against churning. But we don't have the other components (yet). 21:00:57 What's the actual takeaway for users currently though, after reading this paper 21:01:42 Try to match the spending distribution usually, but also add some spicey unpredictability just for fun, I suppose 21:02:11 well, we thought we'd take a stab at it 21:02:46 I'm glad this is all written down fwiw 21:02:47 The randomness needed for wait times before a spend would make Monero pretty unusable though. 21:03:16 Practical churn would have to truncate the distribution of its right side because you could potentially draw a value from the Gamma distribution that would require you to wait months to churn. So maybe...yeah what Diego Salazar said 21:03:51 Yeah, but "needed" covers a massive scope of user actions. You would need to know what a recipient is doing in order to most efficiently test for that pattern 21:04:25 Maybe this test could be extended to see what truncation point is safe. A few hours? A few days? A week? spackle's proof-of-concept churner truncated at about 3 days IIRC 21:05:11 It's a difficult challenge because there's the built-in assumption of trying to model a specific user's actions, but at a universally applicable level 21:05:14 We started this project kind of seeing if there was something obvious that people missed regarding this whole topic. Unlikely, of course. But it's not on MRL's dime so who cares? 21:05:15 Which was about the 60 percentile of the Gamma distribution IIRC. 21:06:16 Diego Salazar: I really appreciate Cypher Stack researching this. You made progress on the research question IMHO :) 21:08:39 The MAGIC Monero Fund also funded research on this, and did not capture Moby: https://monerofund.org/pdf/Borggren-Sept-2023-Probing-the-Attacks-on-the-Privacy-of-the-Monero-Blockchain.pdf 21:09:11 I'm glad to have the writeup; I know this is something we were very focused on having trying to answer in 2018 or so when rings didn't appear to have a near-term deprecation date. To me this still confirms "we don't really have a universal takeaway on churning at this time". Except that ring sigs are bad of course, and bad ring sampling makes them even worse 21:09:44 the MMF research also had the ground truth transaction samples which have been used in other papers as well 21:10:03 oh wait, no not that one. The ACK-J one 21:10:32 I agree the Borggren one also didn't capture Moby, certainly not :) 21:11:10 The ACK-J research: https://raw.githubusercontent.com/ACK-J/Monero-Dataset-Pipeline/main/Lord_of_the_Rings__An_Empirical_Analysis_of_Monero_s_Ring_Signature_Resilience_to_Artificially_Intelligent_Attacks.pdf 21:11:49 the TRM paper used transactions made for that paper 21:12:13 yes, but MRL does not move at the speed of light and CS can't always be of help here while we wait. 21:19:13 Luckily, I know there's a lot that Luke wants your help with in the immediate weeks/months 21:19:33 We're lone wolves. We don't play by other people's rules. 21:19:41 🐺 21:20:20 haha, I'm glad some work was initiated outside of there, that's healthy