network privacy and mixnet §
research §
- Used bounds and asymptotic analysis to create a new version of the table (see this for reference) which estimates the number of layers needed in the mixnet protecting the network layer. The latter assumes a very simple lottery function where the probability of winning is equal to the relative stake.
development §
testnet §
development §
- We are experiencing an issue with the file system on
metal-01.he-eu-hel1.nomos.misc entering a read-only state. This is preventing any write operations on the system, impacting normal operations and services.
- Nomos node stability investigation - found a main reason for tesnet failing after some time, consensus engine has some asserts that are not met during the run. I’m trying to replicate the same behaviour in integration tests.
private PoS §
research §
- A consensus reference page has been written detailing the path and the history of decisions made as a context to the other documents we have provided. It is important to mention that this is a living document that might be updated further in the future.
- Updated the ongoing document of DarkFi Crypsinous implementation with new details.
- Summarized the findings and ideas from different papers around byzantine gossip and consensus. There are some ideas that might be useful to us in the future.
- For the Crypsinous lottery function analyzed the maximum likelihood (ML) statistical inference framework. The latter can be used by an adversary for inference of stakes of nodes to infer relative stake. Formally the ML framework suggests that observations of election statistics of all nodes are required in order to infer relative stake of a single node. However, for long time observations there is a possibility that statistics for a single node are sufficient to infer its relative stake. The analysis of this scenario, which uses a “naive” estimator of relative stake, is currently in progress.
- Normalize stake for leader lottery: Summary with plots and equations can be found here.
- We have a new analytical tool for analyzing the average of our process for learning stake, it lets us study how the process converges without the slow simulation runs and makes analytical work much simpler.
- We found that our process for learning D will underestimate true total stake by up to 3% (depends on our choice of constants), we have some analytical bounds on the underestimate and confirmed them with simulation.
data availability §
research §
- Finished the comparison article of DA structures (Merkle Tree, RS+KZG and Merkle Tree+Snarks) and reach the conclusion that the structure we designed is flexible. We can easily switch to the Merkle tree + Plonky2 structure in the future.
- Some additional research notes in terms of DA comparison can be found here.
- Skimming through NOTRY: Deniable messaging with retroactive avowal to check if there is anything interesting to us
development §
miscellaneous §
- Architecture whitepaper will be reviewed internally by the team in the following weeks.