We have updated our github repository

https://github.com/ConsensusResearch/ForgingSimulation with a new version of the PoS simulation haskell code.

It now included two branches, master - for the single branch classical Nxt based code and "multibranch-experimental" - for the multibranch forging simulation. Recently

new algorithm for regulating tails switching effect is proposed and implemented. With it, a possibility of the N@S attack becomes also regulated as we now can introduce deducible parameter of confirmations needed to stabilize recent blocks tails. The idea of regulating is straightforward - from time to time the node "forgets" almost all the branches and prolong only those whose cumulativeDifficulty measure is above some retargeting threshold. This threshold changes discretely, starting from 0. Unlike the Bitcoin difficulty param, the threshold always grows as the best block cumulativeDifficulty exceeds the previous threshold+delta. So nodes work as multibranch almost all the time, but sometimes becomes "single-branch" for a short time (one tick). This approach allows to have all the multibranch benefits and also get the network with regulating convergence. With a certain confirmation number calculated, we can propose the strong resistance to the N@S as the long tails switching become very-very unlikely after the confirmations. We'll present the N@S simulation results ASAP.

There are more possible regulation procedures, for sure. Basing on the idea that sometimes nodes switch to the single-branch behavior one can introduce any verifiable quasi-random algorithm to do this. The proposed is the simple but efficient one, however more complicated algos (e.g. based on some nice hashes) could secure the system more likely.