[antirez]

Nope. Try yourself with a, b, c, d, e, f, g as guesses. You will see that green letters that are coincident will be the same. So to reconstruct the original SHA256 of the password is easy. The problem then turns like every other hash -> password reconstruction: hard if the original secret is hard to guess via dictionary/brute-force, otherwise easy.

[MattPalmer1086]

Ah, I misunderstood the point you were making. It's still true that each hash won't help you make the next password guess, but you can iteratively fill in parts of the overall hash.

I'm not sure that really helps you much though, as you don't have enough guesses to get the entire hash. And even with that, you may or may not succeed.

Still, good point!

[cestith]

You just need a rainbow table of... 14 character... random passwords... across the allowed symbols. Should be able to build that with Cuda, OpenCL, or OpenMPI in a matter of X weeks given Y hardware budget. Sorry, solving for X and Y is left as an exercise for the reader.

[MattPalmer1086]

Replying to self, if the password is based on a dictionary word, then it's much more doable, as you almost certainly don't need the entire hash. I think you made that point too...