Re:Web's random numbers are too weak, researchers warn
by Tomte on Sunday August 09, 2015 @10:15AM
(#10030333)
So you don't trust modern block ciphers and avoid encrypting more than 128 (or 256) bits of data with a single key?
Isn't that extreme key rotation a bit bothersome?
No, stretching your seed of 256 bits into terabytes of pseudorandom numbers is normal and absolutely fine.
But the pool does not stay with only 256 bits for long (if at all). It's always accumulating more.
Anyway, if the pool ever get to zero, it means that an attacker with infinite resources that can see the entire sequence generated by the CSPRNG could predict the next numbers it'll generate. On practice none of those conditions are met.