15:32:24 It occurs to me that we do not actually need the intra ring randomness. We could have a function that's always gamma like, just with an offset/seed. 15:32:49 That would essentially make every ring encoded as 4 or 5 bytes. 15:33:25 We'd just twiddle the offset so the generated curve hits the real output. 15:34:12 However, this sounds too good to be true. What stupid thing am I doing ? 15:35:12 (beyond leaking some info about the rough age of the real spend, which is something that can be addressed by making offsets larger too for more recent spends) 15:39:47 Do you mind explaining what that would mean for transaction efficiency/new things able to be done with transactions, etc? 15:39:53 I'd also love to get more eyes on: https://github.com/monero-project/research-lab/issues/84 15:40:51 Was that question for me ? 15:41:22 Yes, not sure the implications of what you're suggesting with the offset/seed function 15:41:37 It's a bit over my head but I'd like to understand it 😅 15:41:40 If so, it'd mean smaller txes, since rings are > 4 or 5 bytes now. It'd be especially good with larger rings. 15:41:55 Ok, that's what I figured! 15:42:00 A bit of circular explanation but there's nohting more to it really. 16:06:39 This is the sort of thing proposed by Chator and Green: https://ieeexplore.ieee.org/abstract/document/8406559 16:07:54 PDF: https://isi.jhu.edu/~mgreen/mixing.pdf 16:10:55 Ah, I didn't realize they just removed the function too. Thanks. 16:21:41 ArticMine, luigi1111w: can you give voice to h4sh3d? 16:56:17 I think leaking the rough age of the real spend is a big deal, and should be carefully considered if privacy is simply being reduced for saving space 16:58:30 Alice buys a religious text from Bob in a country where it was illegal. Alice uses an output that is aged long ago, early enough where fewer transactions occurred, or perhaps the age of the output can be correlated to some other piece of info, perhaps her receiving an inheritance or some other financial windfall 17:00:56 The age of the output helps with probabilistic guesses on the purchaser of the illegal religious text 17:04:44 But maybe my concerns are trivial to address :D 18:23:50 moneromooo: regional bins (random selection in clumps) are useful in case your gamma distribution isn’t perfect, or in case an adversary has specific timing data about your tx. Adding more members along the gamma distribution has diminishing returns as well. 18:33:42 just to clarify, the idea is to select ring members around the real output instead of the whole blockchain? 18:35:51 You can’t offset a gamma distribution directly. I.e. if you have a gamma distribution and a real index, if you randomly select a member of the distribution and offset all members so that one lands in the real index, then the resulting members will no longer be on a gamma distribution that matches against the true selection distribution (probability profile of a real spend). This lets you directly analyze the true 18:35:51 spend’s index. 18:37:01 gingeropolous: the idea is to select eg 11 bins from the gamma distribution, and in each bin are eg 11 ring members selected from a small region of the chain 18:38:53 so total ringsize would be 11^11? sorry if im being a dumbass 18:38:58 That crowd paper has the idea to randomly select from a uniform distribution, put the offset in that uniform distribution so the offset is also uniform, then map the distribution onto a gamma distribution. 18:39:04 Yes 18:39:08 Wicked! Better than zcash. 18:39:10 No 18:39:17 Yes sorry 18:39:19 Unless you meant xor 18:39:30 11^2 18:39:41 A bit fuzzy headed 18:41:58 This presentation is helpful http://ieee-sb2018.cs.ucl.ac.uk/presentations/4-Chator.pdf 18:42:18 yeah meant 11x11. sorry 18:42:53 You can set whatever parameters you want; they don't have to be the same 18:45:36 My proposal has floor(sqrt(ring size)) as number of bins, and ring members are distributed among the bins