roadmap/2023-10-09.md at 6b32ce397902979690a0dce3814228f76219daa6

title

`network privacy and mixnet:`

Derived asymptotic expressions for anonymity and communication failure probabilities, when taking into account certain values for population size and network size. Used simulations and analytic framework to study failure probabilities in mixnets of different sizes. Assuming delays between sending and receiving messages are independent random variables from the exponential distribution, derived probability distribution of latency. Main result: one cannot have both low probability of communication and low probability of anonymity failures. The probability of anonymity failure is decreasing with the number of layers but at the expense of increasing the latency.
Finalize research of network-level privacy solutions. Learned important information on: Framework for formalizing privacy, Nym and tokenomics of a Mixnet, Sphinx, and Loopix. Notes found at: https://www.notion.so/Network-Privacy-2dabf0aa878744e299b2ebb97120876f (summaries still WIP).

Lock Overwatch to a specific revision in nomos: https://github.com/logos-co/nomos-node/pull/455.
Implement and pipe services lifecycle handling in Overwatch: https://github.com/logos-co/Overwatch/pull/27.
Bash is being replaced with python due to adding mixnet to docker. At the moment, small issues with node spawning order are being resolved (for tree overlay). ETA on finishing: beginning of this week.
Add API to return mempool item status: https://github.com/logos-co/nomos-node/pull/449.
Make libp2p gossipsub settings configurable: https://github.com/logos-co/nomos-node/pull/454.
Avoid temporary gossipsub errors when bootstrapping tests: https://github.com/logos-co/nomos-node/pull/442.

Exploring multi-staking PPoS design for Carnot.
Idea: a slightly modified version of PoS, unknown how much funds a single validator has and validators are grouped by the amount of stake they have. This property gives validators group-based k-anonymity, where they are indistinguishable (on the stake level). This also enables us to assign the same voting power per each group of validators, which then can be reflected on the overlay structure.
Considering a couple of additional stake hiding/obfuscating techniques without making the initial design too complex. We incentivize/penalize validators based not on the voting power they represent but on the stake. We can have a system that diverges from an equality between voting power and stake, to a system that approximates the voting power based on the stake but the rewarding/penalization directly follows the stake.