6 ms·

What do you mean?

• simias 13y ago
  The output of any good cipher in "counter" mode should be very close to random noise while still being perfectly predictable. Basically any stream cipher worth its salt would pass the parent's test.
  • caf 13y ago
    That's not entirely true for counter mode. For random noise, if you see 2^(n/2) n-bit blocks then there's a ~50% probability of two of those blocks being identical, but if you have that many n-bit blocks from an n-bit block cipher in counter mode the probability of two identical blocks is zero. This doesn't apply to stream ciphers.
    • simias 13y ago
      Interesting but I don't understand why the probability of two identical blocks in this situation would be 0 for CTR mode. As far as I know theoretically there could be a collision and the cipher could output the same value for two different counter values, unless the block cipher guarantees that such a collision can never occur? I was not aware of such a guarantee.
      • caf 13y ago
        A block cipher cannot produce the same output block for two distinct input blocks because it is reversible.
        • acqq 13y ago
          So "just" an order of 2 to 128 block samples has to be collected to see that it's generated? Good luck with that.
          • caf 13y ago
            Nope, AES has a 128-bit block size so you "only" need about 2^64 blocks.
• rainforest 13y ago
  I think the parent means that you can't distinguish a biased RNG from an unbiased one if the bias is designed to look random. In a similar fashion, it's not possible to distinguish cipher text from noise - this impossibility presumably means detecting a biased RNG (i.e. output from an encryption algorithm with known key) is also impossible.