[throwaway7767]

>> A commitment scheme lets you keep a piece of data secret but commit to it so that you can not change it later. >> commitment = SHA256( binding_factor || data ) >> Tell someone the commitment, then [later] reveal both the data and blinding factor.

>It looks like I can change my data, then generate a binding factor that will combine to produce the original hash input.

If you can find SHA256 collisions on demand. But if you can do that, you should probably be writing a paper about it and advancing the state of the art.

[deckar01]

Assume for a moment that no hashing is performed.

I compute C = B || D.

I reveal C.

I later choose new data D'.

I compute C = B' || D'.

I reveal B' and D'.

Since both B and D were secret, B' and D' are accepted.

Secretly masking data lends to malleability. (EDIT: Not a mask)

EDIT: As CJefferson points out the operation is not a mask, but concatenation of a fixed length random value which invalidates this example. Exploiting this secrecy would require a weakness in SHA256 that allows input prefixes to produce colliding hash states (hard).

[CJefferson]

Here || denotes concatenation. Therefore your only options are to change where you split B and D into two two strings. If you (as is common) either fix the length of B, or make sure the splitter marker characters can't occur in B, then given C, B and D are fixed.