[Cryptography Team] Re: block-based cipher padding

[Ron Teitelbaum]

> From: Chris Muller
> Sent: Tuesday, July 11, 2006 1:39 PM
> 
> > The tests that I'm running are only one sided tests, they say encrypt
> > this
> > and your answer should be this.  Since the encryption test didn't
> > assume any
> > padding the results are different, but only by the padding itself.
> 
> Ah, I thought the encryption test DID use some kind of padding.  So if
> it isn't using padding you probably don't want to use our code that
> does padding, right?
> 
> IOW, this sounds like a lower-level test of just the underlying AES w/
> CBC (using three blocks) rather than a test of encryption of an
> arbitrary-sized ByteArray.  Arbitrary-sized ByteArray's are
> encrypted/decrypted with the #encrypt:/#decrypt: methods, respectively.
>  But you just want to encrypt three blocks so you need to use
> #encryptBlock: and #decryptBlock.  They're the lower-level
> functionality that will match the test you're doing, they don't do any
> padding.

The tests do not use padding but do have multiple blocks.  It doesn't seem
wise to break the blocks myself for testing, so I used encrypt: , I guess I
could have used the encrypt:from:to: method but it seems redundant.  Maybe
we should have a encrypt:usePadding: method?  I did like the idea of
abstracting the padding methods to allow for their replacement by
developers.  The encrypt: method could use your padding by default but then
we could have a noPadding option.

BlockCipherPaddingAbstract
	BlockCipherPaddingNoPadding
	BlockCipherPaddingFergusonSchneier2
	...

It would make it much easier to add new padding schemes later.

What do you think?

Ron Teitelbaum