# Re: Complex Theoretical One Way Hash Question

• From: none <""mike\"@(none)">
• Date: Tue, 18 Apr 2006 15:44:44 +0100

ink wrote:
"Matthew Harrison" wrote > Sebastian Gottschalk wrote:

2. Is it feasible ?

No, as it would allow to create a collision, violating the goal of a
cryptographic hash.

You must have misunderstood what I was asking ... I'm not asking for a hash that creates the same result from two different images.

I'm looking at creating a one way hash as normal (with minimal collisions), but ensuring that the resulting hash result is that of the modified image.

<SNIP>

/delurk

Sorry, but isn't that the snake biting its own tail? You can't
put the hash code on the image canvas unless you calculate
it, and you can't calculate it unless you put it on the canvas
first...

Cheers,
ink

/lurk

Well, you could put *any* number in man-readable form in the image, and then make trivial (i.e. visually undetectable) changes to the image until the hash matched the number you chose.
Another question : take the MD5 hash of the phrase " the MD5 hash of this phrase is n", where n is a string of hex digits. For what number n is this true? Birthday paradox, should find one in about 2^64 tries...
.