[klyrs]

Simple answer: of course there is. The real challenge is finding such strings.

If I wanted to know the answer to that question, I'd first search some precomputed rainbow tables to see if anybody's gotten lucky:

http://project-rainbowcrack.com/table.htm

[mdaniel]

The discussion on security.SE seems to imply it's unknown: https://security.stackexchange.com/questions/107081/does-eve... so if you have a proof of your simple answer, I'm sure they'd welcome it

[klyrs]

My statement is one of statistically-informed confidence. A physicist's proof, not a mathematician's. As I said, proving it is where the challenge lies.

[marto1]

I believe proof-by-induction is still a mathematically valid proof. Just one to be suspicious of/can be overridden by more rigid proofs.

[klyrs]

Induction is a mathematical tool. Overwhelming statistics is the physicist's.

[hnfong]

The linked answer on SE basically says statistically there's a near-100% chance that there's a string where md5(string) is 0xfffffff... but there's no concrete proof.

Which is basically the same thing the GP says.

[drdeca]

Do you mean just that it is very very likely that there is, or is there some argument that the MD5 function must be surjective?

[voldacar]

It is not known to be surjective.