From: Eric Biggers <ebiggers@kernel.org>
To: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: "open list:HARDWARE RANDOM NUMBER GENERATOR CORE"
<linux-crypto@vger.kernel.org>,
linux-fscrypt@vger.kernel.org,
linux-arm-kernel <linux-arm-kernel@lists.infradead.org>,
Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
Herbert Xu <herbert@gondor.apana.org.au>,
Paul Crowley <paulcrowley@google.com>,
Greg Kaiser <gkaiser@google.com>,
Michael Halcrow <mhalcrow@google.com>,
"Jason A . Donenfeld" <Jason@zx2c4.com>,
Samuel Neves <samuel.c.p.neves@gmail.com>,
Tomer Ashur <tomer.ashur@esat.kuleuven.be>
Subject: Re: [RFC PATCH v2 11/12] crypto: adiantum - add Adiantum support
Date: Wed, 24 Oct 2018 15:06:17 -0700 [thread overview]
Message-ID: <20181024220614.GA13497@gmail.com> (raw)
In-Reply-To: <CAKv+Gu9zBKK7ughbBAKkHdfgX2KrBXkF6MhH3xrSdLegrjjYJQ@mail.gmail.com>
On Tue, Oct 23, 2018 at 07:40:34AM -0300, Ard Biesheuvel wrote:
> On 20 October 2018 at 15:12, Eric Biggers <ebiggers@kernel.org> wrote:
> > Hi Ard,
> >
> > On Sat, Oct 20, 2018 at 12:17:58PM +0800, Ard Biesheuvel wrote:
> >> On 16 October 2018 at 01:54, Eric Biggers <ebiggers@kernel.org> wrote:
> >> > From: Eric Biggers <ebiggers@google.com>
> >> >
> >> > Add support for the Adiantum encryption mode. Adiantum was designed by
> >> > Paul Crowley and is specified by our paper:
> >> >
> >> > Adiantum: length-preserving encryption for entry-level processors
> >> > (https://eprint.iacr.org/2018/720.pdf)
> >> >
> >> > See our paper for full details; this patch only provides an overview.
> >> >
> >> > Adiantum is a tweakable, length-preserving encryption mode designed for
> >> > fast and secure disk encryption, especially on CPUs without dedicated
> >> > crypto instructions. Adiantum encrypts each sector using the XChaCha12
> >> > stream cipher, two passes of an ε-almost-∆-universal (εA∆U) hash
> >> > function, and an invocation of the AES-256 block cipher on a single
> >> > 16-byte block. On CPUs without AES instructions, Adiantum is much
> >> > faster than AES-XTS; for example, on ARM Cortex-A7, on 4096-byte sectors
> >> > Adiantum encryption is about 4 times faster than AES-256-XTS encryption,
> >> > and decryption about 5 times faster.
> >> >
> >> > Adiantum is a specialization of the more general HBSH construction. Our
> >> > earlier proposal, HPolyC, was also a HBSH specialization, but it used a
> >> > different εA∆U hash function, one based on Poly1305 only. Adiantum's
> >> > εA∆U hash function, which is based primarily on the "NH" hash function
> >> > like that used in UMAC (RFC4418), is about twice as fast as HPolyC's;
> >> > consequently, Adiantum is about 20% faster than HPolyC.
> >> >
> >> > This speed comes with no loss of security: Adiantum is provably just as
> >> > secure as HPolyC, in fact slightly *more* secure. Like HPolyC,
> >> > Adiantum's security is reducible to that of XChaCha12 and AES-256,
> >> > subject to a security bound. XChaCha12 itself has a security reduction
> >> > to ChaCha12. Therefore, one need not "trust" Adiantum; one need only
> >> > trust ChaCha12 and AES-256. Note that the εA∆U hash function is only
> >> > used for its proven combinatorical properties so cannot be "broken".
> >> >
> >>
> >> So what happens if the part of the input covered by the block cipher
> >> is identical between different generations of the same disk block
> >> (whose sector count is used as the 'outer' IV). How are we not in the
> >> same boat as before when using stream ciphers for disk encryption?
> >>
> >
> > This is the point of the hash step. The value encrypted with the block cipher
> > to produce the intermediate value C_M (used as the stream cipher nonce) is
> > H(T, P_L) + P_R. (T is the tweak a.k.a the IV, P_L is the plaintext except the
> > last 16 bytes, P_R is the last 16 bytes.) A collision in this value occurs iff:
> >
> > H(T1, P1_L) + P1_R = H(T2, P2_L) + P2_R
> > i.e.
> > H(T1, P1_L) - H(T2, P2_L) = P2_R - P1_R
> >
> > If (T1, P1_L) = (T2, P2_L) then P1_R != P2_R so the equation has no solutions
> > (since we don't consider queries where the whole input is the same; those
> > unavoidably produce the same ciphertext). Otherwise (T1, P1_L) != (T2, P2_L),
> > and since the hash function H is ε-almost-∆-universal over integers mod 2^128,
> > the equation is true for at most a very small proportion 'ε' of hash keys.
> > But, the hash key is chosen at random and is unknown to the attacker.
> >
> > The same applies in the other direction, for chosen ciphertext attacks.
> >
> > Basically, it's very difficult for an attacker to cause the intermediate value
> > C_M to be reused, and the outputs will appear random until they do.
> >
> > Of course, all this is explained much more precisely and comprehensively in our
> > paper. See section 5, "Security reduction".
> >
>
> Thanks for the explanation. I saw that the result of the AES
> encryption was used as the XChaCha nonce, but I failed to spot that
> the result of the nhpoly1305 pass is added/subtracted to/from that
> particular block first.
>
> In any case, this looks good to me: as far as I can tell, the code
> implements the algorithm as described in the paper, and the plumbing
> into the crypto API looks correct to me as well.
>
> Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
>
> Whether the paper is correct is a different matter: it looks
> convincing to me but IANAC.
>
> The only request I have is to add a speed test to tcrypt as well so we
> can easily benchmark it.
I'll add an Adiantum testing mode to tcrypt, though I have to admit that I
really dislike tcrypt, so I've actually been using a custom patch instead
(https://git.kernel.org/pub/scm/linux/kernel/git/ebiggers/linux.git/commit/?h=cryptobench).
FWIW, I mostly dislike tcrypt's lack of flexibility. Every benchmark setting
has to be coded explicitly in tcrypt, with a mode number to select it. It's
often missing what I want, or what I want is included but comes with a bunch of
unwanted tests too. Also tcrypt has to be built as a loadable module, so it
can't be used with CONFIG_MODULES=n. And the benchmark results only go directly
to the kernel log, where they aren't easily retrievable from a script.
The patch I've been using just exposes a file /proc/cryptobench (note: it maybe
should be made into a char device instead) to which you can write commands like:
algtype=skcipher algname=adiantum(xchacha12,aes) keysize=32 bufsize=4096 niter=1000
Then userspace reads back the result:
SUCCESS algname=adiantum(xchacha12,aes) driver_name=adiantum(xchacha12-software,aes-aesni,nhpoly1305-generic) measurement=0x30a5abd5776e0af enc_time=44831104 dec_time=38303077
It's then pretty straightforward to wrap this API in a userspace script that
does all the benchmarks you need, and formats the results nicely.
For correctness testing there's also an option 'sgl_fuzz' that randomizes the
scatterlist division.
Currently the patch is missing some features such as AEAD support, but at some
point I'd like to get it into a state where it can be included upstream.
- Eric
next prev parent reply other threads:[~2018-10-24 22:06 UTC|newest]
Thread overview: 54+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-10-15 17:54 [RFC PATCH v2 00/12] crypto: Adiantum support Eric Biggers
2018-10-15 17:54 ` [RFC PATCH v2 01/12] crypto: chacha20-generic - add HChaCha20 library function Eric Biggers
2018-10-19 14:13 ` Ard Biesheuvel
2018-10-15 17:54 ` [RFC PATCH v2 02/12] crypto: chacha20-generic - add XChaCha20 support Eric Biggers
2018-10-19 14:24 ` Ard Biesheuvel
2018-10-15 17:54 ` [RFC PATCH v2 03/12] crypto: chacha20-generic - refactor to allow varying number of rounds Eric Biggers
2018-10-19 14:25 ` Ard Biesheuvel
2018-10-15 17:54 ` [RFC PATCH v2 04/12] crypto: chacha - add XChaCha12 support Eric Biggers
2018-10-19 14:34 ` Ard Biesheuvel
2018-10-19 18:28 ` Eric Biggers
2018-10-15 17:54 ` [RFC PATCH v2 05/12] crypto: arm/chacha20 - add XChaCha20 support Eric Biggers
2018-10-20 2:29 ` Ard Biesheuvel
2018-10-15 17:54 ` [RFC PATCH v2 06/12] crypto: arm/chacha20 - refactor to allow varying number of rounds Eric Biggers
2018-10-20 3:35 ` Ard Biesheuvel
2018-10-20 5:26 ` Eric Biggers
2018-10-15 17:54 ` [RFC PATCH v2 07/12] crypto: arm/chacha - add XChaCha12 support Eric Biggers
2018-10-20 3:36 ` Ard Biesheuvel
2018-10-15 17:54 ` [RFC PATCH v2 08/12] crypto: poly1305 - add Poly1305 core API Eric Biggers
2018-10-20 3:45 ` Ard Biesheuvel
2018-10-15 17:54 ` [RFC PATCH v2 09/12] crypto: nhpoly1305 - add NHPoly1305 support Eric Biggers
2018-10-20 4:00 ` Ard Biesheuvel
2018-10-20 5:38 ` Eric Biggers
2018-10-20 15:06 ` Ard Biesheuvel
2018-10-22 18:42 ` Eric Biggers
2018-10-22 22:25 ` Ard Biesheuvel
2018-10-22 22:40 ` Eric Biggers
2018-10-22 22:43 ` Ard Biesheuvel
2018-10-15 17:54 ` [RFC PATCH v2 10/12] crypto: arm/nhpoly1305 - add NEON-accelerated NHPoly1305 Eric Biggers
2018-10-20 4:12 ` Ard Biesheuvel
2018-10-20 5:51 ` Eric Biggers
2018-10-20 15:00 ` Ard Biesheuvel
2018-10-15 17:54 ` [RFC PATCH v2 11/12] crypto: adiantum - add Adiantum support Eric Biggers
2018-10-20 4:17 ` Ard Biesheuvel
2018-10-20 7:12 ` Eric Biggers
2018-10-23 10:40 ` Ard Biesheuvel
2018-10-24 22:06 ` Eric Biggers [this message]
2018-10-30 8:17 ` Herbert Xu
2018-10-15 17:54 ` [RFC PATCH v2 12/12] fscrypt: " Eric Biggers
2018-10-19 15:58 ` [RFC PATCH v2 00/12] crypto: " Jason A. Donenfeld
2018-10-19 18:19 ` Paul Crowley
2018-10-20 3:24 ` Ard Biesheuvel
2018-10-20 5:22 ` Eric Biggers
[not found] ` <2395454e-a0dc-408f-4138-9d15ab5f20b8@esat.kuleuven.be>
2018-10-22 11:20 ` Tomer Ashur
2018-10-19 19:04 ` Eric Biggers
2018-10-20 10:26 ` Milan Broz
2018-10-20 13:47 ` Jason A. Donenfeld
2018-11-16 21:52 ` Eric Biggers
2018-11-17 10:29 ` Milan Broz
2018-11-19 19:28 ` Eric Biggers
2018-11-19 20:05 ` Milan Broz
2018-11-19 20:30 ` Jason A. Donenfeld
2018-10-21 22:23 ` Eric Biggers
2018-10-21 22:51 ` Jason A. Donenfeld
2018-10-22 17:17 ` Paul Crowley
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20181024220614.GA13497@gmail.com \
--to=ebiggers@kernel.org \
--cc=Jason@zx2c4.com \
--cc=ard.biesheuvel@linaro.org \
--cc=gkaiser@google.com \
--cc=herbert@gondor.apana.org.au \
--cc=linux-arm-kernel@lists.infradead.org \
--cc=linux-crypto@vger.kernel.org \
--cc=linux-fscrypt@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=mhalcrow@google.com \
--cc=paulcrowley@google.com \
--cc=samuel.c.p.neves@gmail.com \
--cc=tomer.ashur@esat.kuleuven.be \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).