Re: [RFC PATCH v2] md/dm-crypt - reuse eboiv skcipher for IV generation

[Eric Biggers <ebiggers@kernel.org>]

On Thu, Aug 08, 2019 at 01:23:10PM +0000, Pascal Van Leeuwen wrote:
> > -----Original Message-----
> > From: Milan Broz <gmazyland@gmail.com>
> > Sent: Thursday, August 8, 2019 2:53 PM
> > To: Pascal Van Leeuwen <pvanleeuwen@verimatrix.com>; Eric Biggers <ebiggers@kernel.org>
> > Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>; linux-crypto@vger.kernel.org;
> > herbert@gondor.apana.org.au; agk@redhat.com; snitzer@redhat.com; dm-devel@redhat.com
> > Subject: Re: [RFC PATCH v2] md/dm-crypt - reuse eboiv skcipher for IV generation
> > 
> > On 08/08/2019 11:31, Pascal Van Leeuwen wrote:
> > >> -----Original Message-----
> > >> From: Eric Biggers <ebiggers@kernel.org>
> > >> Sent: Thursday, August 8, 2019 10:31 AM
> > >> To: Pascal Van Leeuwen <pvanleeuwen@verimatrix.com>
> > >> Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>; linux-crypto@vger.kernel.org;
> > >> herbert@gondor.apana.org.au; agk@redhat.com; snitzer@redhat.com; dm-devel@redhat.com;
> > >> gmazyland@gmail.com
> > >> Subject: Re: [RFC PATCH v2] md/dm-crypt - reuse eboiv skcipher for IV generation
> > >>
> > >> On Wed, Aug 07, 2019 at 04:14:22PM +0000, Pascal Van Leeuwen wrote:
> > >>>>>> In your case, we are not dealing with known plaintext attacks,
> > >>>>>>
> > >>>>> Since this is XTS, which is used for disk encryption, I would argue
> > >>>>> we do! For the tweak encryption, the sector number is known plaintext,
> > >>>>> same as for EBOIV. Also, you may be able to control data being written
> > >>>>> to the disk encrypted, either directly or indirectly.
> > >>>>> OK, part of the data into the CTS encryption will be previous ciphertext,
> > >>>>> but that may be just 1 byte with the rest being the known plaintext.
> > >>>>>
> > >>>>
> > >>>> The tweak encryption uses a dedicated key, so leaking it does not have
> > >>>> the same impact as it does in the EBOIV case.
> > >>>>
> > >>> Well ... yes and no. The spec defines them as seperately controllable -
> > >>> deviating from the original XEX definition - but in most practicle use cases
> > >>> I've seen, the same key is used for both, as having 2 keys just increases
> > >>> key  storage requirements and does not actually improve effective security
> > >>> (of the algorithm itself, implementation peculiarities like this one aside
> > >>> :-), as  XEX has been proven secure using a single key. And the security
> > >>> proof for XTS actually builds on that while using 2 keys deviates from it.
> > >>>
> > >>
> > >> This is a common misconception.  Actually, XTS needs 2 distinct keys to be a
> > >> CCA-secure tweakable block cipher, due to another subtle difference from XEX:
> > >> XEX (by which I really mean "XEX[E,2]") builds the sequence of masks starting
> > >> with x^1, while XTS starts with x^0.  If only 1 key is used, the inclusion of
> > >> the 0th power in XTS allows the attack described in Section 6 of the XEX paper
> > >> (https://web.cs.ucdavis.edu/~rogaway/papers/offsets.pdf).
> > >>
> > > Interesting ... I'm not a cryptographer, just a humble HW engineer specialized
> > > in implementing crypto. I'm basing my views mostly on the Liskov/Minematsu
> > > "Comments on XTS", who assert that using 2 keys in XTS was misguided.
> > > (and I never saw any follow-on comments asserting that this view was wrong ...)
> > > On not avoiding j=0 in the XTS spec they actually comment:
> > > "This difference is significant in security, but has no impact on effectiveness
> > > for practical applications.", which I read as "not relevant for normal use".

See page 6 of "Comments on XTS":

	Note that j = 0 must be excluded, as f(0, v) = v for any v, which
	implies ρ = 1. Moreover, if j = 0 was allowed, a simple attack based on
	this fact existed, as pointed out by [6] and [3]. Hence if XEX is used,
	one must be careful to avoid j being 0.

The part you quoted is only talking about XTS *as specified*, i.e. with 2 keys.

> > >
> > > In any case, it's frequently *used* with both keys being equal for performance
> > > and key storage reasons.

It's broken, so it's broken.  Doesn't matter who is using it.

> > 
> > There is already check in kernel for XTS "weak" keys (tweak and encryption keys must not be
> > the same).
> > 
> > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/include/crypto/xts.h#
> > n27
> > 
> > For now it applies only in FIPS mode... (and if I see correctly it is duplicated in all
> > drivers).
> > 
> I never had any need to look into FIPS for XTS before, but this actually appears
> to be accurate. FIPS indeed *requires this*. Much to my surprise, I might add.
> Still looking for some actual rationale that goes beyond suggestion and innuendo 
> (and is not too heavy on the math ;-) though.

As I said, the attack is explained in the original XEX paper.  Basically the
adversary can submit a chosen ciphertext query for the first block of sector 0
to leak the first "mask" of that sector, then submit a chosen plaintext or
ciphertext query for the reminder of the sector such that they can predict the
output with 100% certainty.  (The standard security model for tweakable block
ciphers says the output must appear random.)

- Eric