--- layout: post title: Logs for the Monero Research Lab Meeting Held on 2019-09-16 summary: Sarang work, and miscellaneous tags: [dev diaries, crypto, research] author: el00ruobuob / sarang --- # Logs **\** OK, it's time for the meeting! **\** Agenda: https://github.com/monero-project/meta/issues/390 **\** Logs posted there afterward **\** GREETINGS **\** Hello **\** I'll wait a couple of minutes in case anyone else shows up **\** o/ **\** \*the regular crowd shuffles in\* **\** Hi **\** Our pal suraeNoether said he may not be available for today's meeting **\** But I can share some of the things I've been working on for our ROUNDTABLE **\** The ever-clever RandomRun posted an idea for a signature scheme earlier: https://github.com/monero-project/research-lab/issues/56 **\** Some updates have been made for efficiency, and I worked up proof-of-concept code: https://github.com/SarangNoether/skunkworks/tree/lrs/lrs **\** And a timing/space analysis: https://github.com/SarangNoether/skunkworks/blob/sublinear/triptych.md **\** (I gave it the name Triptych as a placeholder, so we have a name to use for clarity) **\** It actually beats Lelantus in terms of 2-2 transaction size **\** But verification is less efficient **\** Also note that security hasn't been proven yet, but it uses a modification by Bootle et al. to a 1-of-N proof by Groth **\** and that 1-of-N has good proofs **\** Aside from that, I've been working with the Lelantus authors on some ideas to fix its self-spend tracing problem **\** And that's coming together nicely **\** The CLSAG paper will be submitted to Financial Cryptography this week **\** And my CCS funding request for next quarter has been opened: https://ccs.getmonero.org/proposals/sarang-2019-q4.html **\** On a more whimsical note, a preprint was just posted that does some analysis on a card-based cipher originally designed by Bruce Schneier for a book: https://arxiv.org/abs/1909.06300 **\** It's a neat example of a cipher that appears to resist a good deal of modern cryptanalysis, but can be done using paper, pen, and a deck of playing cards! **\** ElsieFour also has such properties except without the playing cards. **\** Ah, and I'd be remiss if I didn't mention the trustless recursive SNARK paper, Halo, that was recently posted by the Zcash folks **\** mikerah: I wasn't familiar with that! **\** Has it undergone much analysis? **\** Here's the preprint: https://eprint.iacr.org/2017/339.pdf **\** neat **\** I'm not sure if it has gone through much analysis as it's a relatively new construction. **\** But you can use paper and pen! **\** Halo has some clever ideas in it, but it's worth noting (as usual) that preprints don't undergo peer review, and that Halo currently lacks a soundness proof **\** It will be fun to see the new research that comes from its ideas **\** Any particular questions on the items that I mentioned? **\** How would the ideas from lelantus get implemented in monero? **\** Its transaction model could, hypothetically, be implemented directly **\** Using a particular kind of migration transaction to transition older outputs **\** It would result initially in a smaller anonymity set **\** Currently Lelantus has a tracing issue that's a deal-breaker IMO **\** but very recent ideas mean that may not be a problem **\** Would there be traceability problems from the current monero blockchain to this hypothetical lelantus+monero blockchain? **\** How so? **\** As in, would it be possible to trace transactions between hard forked blockchains **\** In such an implementation, old-style transactions would not be allowed **\** Old outputs would undergo a signer-ambiguous transaction to generate a new output commitment that is Lelantus-compatible **\** sarang: greets **\** So a migration is trivially distinguishable, but retains the same kind of signer ambiguity that exists now **\** hi **\** To be clear, there are no plans to implement this AFAIK **\** I see. I guess more work would need to be done on this front. **\** It's all just research **\** Anyway, that's what I've been working on **\** Does anyone else wish to share interesting research? **\** OK! **\** Well, in that case my ACTION ITEMS are administrative stuff for FC submission, ongoing analysis of Lelantus modifications and proofs, and returning to some existing recent proving systems **\** Before we adjourn, is there anything else to discuss? **\** i don't have any research im working on, but im enjoying banging my head regarding the randomx branch prediction problem **\** Go on! **\** so, big chunk of CPU silicon dedicated to branch prediction. Turns out a lot of the methods use neural networks kinda thing (called perceptron at one point). **\** however, problem is that randomx is random - its random whether a branch will be taken **\** and when somethings random, hard for machine-learning / pattern recognition to get any gains **\** Makes sense **\** however, if you try and seed random into the program (such that a branch predictor could find some emergent pattern), this information could be harvested by an ASIC or some other mitigation **\** so, my head sorta got stuck at that point... and if it'd be possible to somehow hide the emergent pattern... and then all the thought threads frayed **\** So, using information from existing CPU architectures in order to develop better specialized hardware? **\** Or information from any kind of well-designed predictor, I suppose **\** well the general randomx problem is to make a PoW that leverages stuff in CPUs. **\** and branch prediction is underleveraged due to the problem i just described **\** Ah, ok **\** I don't know enough about CPU branch prediction to fully appreciate this, but it sounds interesting nonetheless **\** Anything else of interest to share before the meeting ends? **\** All righty then **\** Thanks to everyone for being here; we are now adjourned! **\** Logs will be posted shortly