February 8th, 2015, 11:28 am
QuoteOriginally posted by: outrunQuoteOriginally posted by: Traden4AlphaCool!So (f(i,j) and f(k,l) initialize completely independent RN streams as long as abs(i-k)+abs(j-l)>0?What about statistical comparisons of f(i,j), f(i+1,j), f(i,j+1)? Do changes in the first input have a different effect compared to changes in the second input?I don't understand the abs? Maybe there is still a bit of misunderstanding?It's just a way of saying that at least one of the input variables is different between the two initializations.QuoteOriginally posted by: outrunSuppose you have the 256 bit counter 0, 1, 2, ..., 2^256-1Next you encrypt each counter number with an encryption key "k", and some encryption function f(). f() can be RSA, EAS,.. etcthis give f(0,k), f(1,k), f(2,k), ..., f(2^256-1,k)and it's a permutations of all the counter value 0 ... 2^256-1. now if you use a different value of k then you'll get a complete different sequence, a different permutation of all legal counter values.Interesting. So, if one created a 512 bit counter and split it into two halves, i and k, one would have an RNG that produces 2^512 RNs drawn from the space of 0 ... 2^256-1. I assume three fish passes all the tests for RNs.