|
Posted by Unruh on December 26, 2007, 1:17 pm
If you were Registered and logged in, you could reply and use other advanced thread options
>Unruh wrote:
>>> Unless you reach its maxixum output length (which is typically very short in
>>> comparison to the input), any good cryptographic hash does its best in
>>> preserving as much information as possible.
>>
>> No, it does its best at not preserving any information as possible if it is
>> a cryptographic hash.
>This is ridiculous. Consider an input of the length of the output 'n' with
>maximum conditional entropy. If the output would contain significantly less
>entropy, say m < n, then the average runtime for a bruteforce search would
>be 2^m instead of 2^n, and you'd have an attack against the hash.
???? A cryptographic hash tries very hard to look like a random selection
of the 2^N possible outputs(for a N bit hash). That preserves no
information. Information is what distinguishes the input from other inputs,
and a cryptographic has tries its best not to distinguish the outputs from
any other outputs. It, it tries its best to destroy all information in the
input. The reason for the "each bit influences the output" is precisely to
make this as true as possible. You also want to make sure that different
inputs produce different outputs. Ie, each input is a statistically
independent random choice of output. If one of the bits did not influence
the output, then you would not have an independent random choice.
Of course the output is actually a deterministic function of the input, but
one wants that deterministic function to act as much like a random
selection as possible-- ie to preserve the "no information transfer" as
possible.
>Short to say, cryptographic hashes are best at mixing in all available
>information without throwing anything away. That is, every little input
>influences the output with maximum significance.
>> Yes, he did. He implied that he was talking about a situation in which is
>> was very hard or impossible to reverse the hash. Or did you not happen to
>> read the OP post.
>If the input is longer than the output, a hash is always impossible to
>invert for arbitrary inputs - that's the very purpose of a hash. No one
>every talked about that it should also be hard for specific inputs.
Yes, you are again repeating what others said.
>> ??? Where did I say you know the key?
>It's part of the algorithm if you use it solely as a hash.
Who was using it as a hash? Using a keyed encryption as a cryptographic hash is
silly
both because it preserves the length of the input, and because it is easily
invertible.
| 
XML site map