I am interested in using the new
PREVRANDAO value as a source of pseudo-randomness in calculating the Miller-Rabin probabilistic primality test on-chain. Chainlink VRF is an option but ideally I would like to not have any external dependencies in my contract.
Architecturally, I would plan to use the value as a seed to generate a sufficient number of
a bases to be used in the test.
My understanding is that it is only pseudo-random, and not truly random, as the validator has the option to choose NOT to propose a block, IF the current randomness value is favourable to them in some context. This differs from eg. using
block.blockhash or others for randomness as the validator has a lot more influence over the value in this case.
My question is, how likely is it for this to occur, given that I generate a sufficient number of
a bases? Given that the value is random the test should not be affected, but the validator still has influence over whether it's used or not.