Continuing the research on the effects of wealth concentration within Ouroboros Crypsinous - more precisely fixing some of the code and running simulations. The results and reports will continue to be provided in the Notion doc.
Also, doing additional research into the Darkfi’s Crypsinous implementation (and why they are striving away from it)
Mixnet specification is done, currently in review - once additional comments have been added, will provide new updates
Studying byzantine gossip and swarm consensus as part of the effort of solving the tagging attack problem (as alternatives to random subsampling and reliable broadcast) - some of the ideas involve the use of gradecast over byzantine gossip, but providing more details in the following week.
By fixing the above and the depth calculation in the analysis script, the simulations stall has been fixed totally, and we are meeting in the deadlines in terms of releasing the Carnot papers - no additional showstoppers
We’ve run simulations of the process and they confirm the stability conditions and convergence derived analytically (above)
There is still work to be done in terms of slots observation to understand a good measurement of leader rate as well as some additional confirmation of the analysis
Used the lower bound (LB), asymptotic estimate (AE) and upper bound (UB) on the probability of stake greater than some threshold (“confidence”), within a given “accuracy”, to estimate the number of layers in the mixnet.
The UB (on the number of layers) is loose. The AE (for the number of layers) is close to the LB (on the number of layers), suggesting that LB is closer to true values (AE is in very good agreement with simulations). However, AE is only available for “confidence” higher than 0.6 and some “accuracies”
For the currently used lottery function, derived the maximum likelihood estimator of relative stake. The latter can be used by an adversary to infer the relative stake of a node
Analysis of fraction of compromised paths is currently in progress and more details will be provided in the following weeks
The parameters we are currently comparing are proof size, prover time, verifier time and commitment size, but in terms of theoretical examples, not actual benchmarks.
development
The new read/write DA API is still in progress - there are conversations ongoing which can be seen in the draft PR
miscellaneous
The architecture whitepaper is awaiting feedback and expected to be finalized in the first week of 2024 - the paper can be found on Discord