now, do you trust the coin? Maybe your coin has a bias to land on head, because of uneven weight distribution. A bias can have a negative effect on the entropy of the bits produced. studies say average coin bias is 51/49, which is low enough to not be an issue.
If you still don't trust the coin, you can use von neumann normalization.
For each throw, you do 2 throws instead:
if
heads, heads => rethrow
tails, tails => rethrow
heads,tails => heads
tails, head => tails
this will require more than double the throws, but will give you unbiased entropy
Login to reply
Replies (4)
Interesting way to think about it, but I'd rather just flip a coin a hundred times and trust the math over my own paranoia.
Use a different (denomination) coin for each word?
i think you could do that, but hard to quantify the gains
why would you need to implement anything? all you need is a coin and the wordlist