Fwd: Cloudflare public comments on FIPS IPD 203, 204, and 205.

640 views
Skip to first unread message

Bas Westerbaan

unread,
Nov 26, 2023, 6:03:18 AM11/26/23
to pqc-forum
We sent in the following comments last week.

Best,

 Bas
Cloudflare comments on FIPS IPD 203, 204, and 205.pdf

Jordi Mur work

unread,
Nov 26, 2023, 12:03:31 PM11/26/23
to pqc-forum, Bas Westerbaan
Hi Bas,

Many thanks for sharing Cloudflare's comments on the FIPS IPDs in this forum.

Could you comment on the range of values on the x-axis (handshake time) graph showing the cumulative distribution function of TLS handshake times of PQ and non-PQ connections? One could make an educated guess based on your 2019 experiments, but it'd be preferable to hear it from the authors.

Best regards,
Jordi

El dia diumenge, 26 de novembre de 2023 a les 12:03:18 UTC+1, Bas Westerbaan va escriure:

Bas Westerbaan

unread,
Nov 26, 2023, 1:07:45 PM11/26/23
to Jordi Mur work, pqc-forum
Could you comment on the range of values on the x-axis (handshake time) graph showing the cumulative distribution function of TLS handshake times of PQ and non-PQ connections?

The axis is logarithmic, but the labels are left out on purpose. Clearing publication of those required a bit more process than we had time for. We might share them in a future publication.

Best,

 Bas

bruno

unread,
Nov 27, 2023, 5:07:33 AM11/27/23
to pqc-...@list.nist.gov

Hi Bas, Thanks for the work

1) Would you mind to provide more inputs in the light of the interesting percentage in the note as it is quiet unclear to me of the impact of the current implementation ", is used by 1.7% of all our inbound TLS 1.3" and "At the time of writing, 7% of all Chrome 118+ TLS 1.3". This is an interesting topic for end-user to understand the fingerprint/measurements of Cloudflare as well as other massive providers...I understand you are writing some paper so that's not urgent but i have not been able to understand the adoption overall.

2) I noticed in the draft RFC (X25519Kyber768Draft00  that you are proposing a concatenation of the 32 bytes shared secrets. It sounds to me not really aligned with NIST SP 800-56C REV. 2 which mandates a KDF of both PQC and non PQC. I guess it will be interesting to clarify this point for compliance and security reasons.

3) In regard of the work on X25519Kyber768Draft00,  I had in mind that QUIC is having some traction so it might be good to understand what could be re-used. I put it here since I have not looked at Quic so far so may be this is remark can be discarded?.

Thanks

Regards

--
You received this message because you are subscribed to the Google Groups "pqc-forum" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pqc-forum+...@list.nist.gov.
To view this discussion on the web visit https://groups.google.com/a/list.nist.gov/d/msgid/pqc-forum/CAMjbhoXFOCTExeDg9mHC_AGUqgvF6ieZ7DD5%3D8N9RtDdRD4H6A%40mail.gmail.com.

Bas Westerbaan

unread,
Nov 27, 2023, 5:33:36 AM11/27/23
to bruno, pqc-...@list.nist.gov

1) Would you mind to provide more inputs in the light of the interesting percentage in the note as it is quiet unclear to me of the impact of the current implementation

Sure, what is unclear to you?

2) I noticed in the draft RFC (X25519Kyber768Draft00  that you are proposing a concatenation of the 32 bytes shared secrets. It sounds to me not really aligned with NIST SP 800-56C REV. 2 which mandates a KDF of both PQC and non PQC. I guess it will be interesting to clarify this point for compliance and security reasons.

See the "hybrid" variety. (Second paragraph of §2 of SP 800-56C rev 2.)

3) In regard of the work on X25519Kyber768Draft00,  I had in mind that QUIC is having some traction so it might be good to understand what could be re-used. I put it here since I have not looked at Quic so far so may be this is remark can be discarded?.

QUIC essentially uses the same key agreement as TLS 1.3: this covers both. The "1.7%" covers both TLS 1.3 over TCP and QUIC.
 
Best,

 Bas

Mike Ounsworth

unread,
Nov 27, 2023, 1:02:46 PM11/27/23
to Bas Westerbaan, bruno, pqc-...@list.nist.gov
>> 2) I noticed in the draft RFC (X25519Kyber768Draft00 that you are proposing a concatenation of the 32 bytes shared secrets. It sounds to me not really aligned with NIST SP 800-56C REV. 2 which mandates a KDF of both PQC and non PQC. I guess it will be interesting to clarify this point for compliance and security reasons.
> See the "hybrid" variety. (Second paragraph of §2 of SP 800-56C rev 2.)

If I understand your construction correctly, there is already assumed to be an HKDF happening in the surrounding HPKE protocol, so doing a KDF inside the X25519Kyber768Draft00 algorithm would be redundant.

Although, as Scott Fluhrer so eloquently put during the last IETF PQUIP; a lot of this is at the discretion of the FIPS lab, so let's not count our chickens until they've been certified. The more obviously it matches SP 800-56Cr2, the less trouble we expect at certification time, even if it's "cryptographically equivalent".

---
Mike Ounsworth
--
You received this message because you are subscribed to the Google Groups "pqc-forum" group.
To unsubscribe from this group and stop receiving emails from it, send an email to mailto:pqc-forum+...@list.nist.gov.
To view this discussion on the web visit https://urldefense.com/v3/__https://groups.google.com/a/list.nist.gov/d/msgid/pqc-forum/CAMjbhoWSwmw_DPZeEzMfXOcbGy61i0tg24Qyb1xkK0aQsrOVzw*40mail.gmail.com?utm_medium=email&utm_source=footer__;JQ!!FJ-Y8qCqXTj2!Zicvkom3eyLVSNJy5IyyVuq33cKGo0NhK8c1nF_je1fsQrGIOE-Mnh7aUi5zXh_Dx2jkgGgkmDXwEmUGUg2av5wr0yXn$.

Sophie Schmieg

unread,
Nov 27, 2023, 2:01:22 PM11/27/23
to Mike Ounsworth, Bas Westerbaan, bruno, pqc-...@list.nist.gov
Note that X25519Kyber768Draft00 predates the publication of the NIST standard and therefore still uses the version of Kyber in Round 3, which includes an additional KDF step as well. TLS will run a KDF on the concatenated secrets and the session transcript in any case, so the current implementation is using two KDFs, but once we switch to NIST's version of Kyber I would presume that only one KDF over the concatenated secrets and session transcript as per usual TLS rules will be used. I don't think adding additional KDF calls would be required for TLS, and consider the round 3 KDF as merely an artifact of the previous algorithm description.

To unsubscribe from this group and stop receiving emails from it, send an email to pqc-forum+...@list.nist.gov.
To view this discussion on the web visit https://groups.google.com/a/list.nist.gov/d/msgid/pqc-forum/CH0PR11MB5739239CCB440E71EE5A31A29FBDA%40CH0PR11MB5739.namprd11.prod.outlook.com.


--

Sophie Schmieg |
 Information Security Engineer | ISE Crypto | ssch...@google.com

Mike Ounsworth

unread,
Nov 27, 2023, 3:57:58 PM11/27/23
to Sophie Schmieg, Bas Westerbaan, bruno, pqc-...@list.nist.gov

Hi Sophie,

 

I’m not a great expert here, but I thought that X25519Kyber768Draft00 was an HPKE algorithm – therefore only applicable to the EncryptedClientHello, and not a general TLS KeyEx algorithm?

 

Genuine question: does the ECH need to be cryptographically self-contained, or is it ok for the security of ECH to depend on the TLS transcript check at the very end?

 

---

Mike Ounsworth

Bas Westerbaan

unread,
Nov 27, 2023, 4:38:54 PM11/27/23
to Mike Ounsworth, Sophie Schmieg, bruno, pqc-...@list.nist.gov
On Mon, Nov 27, 2023 at 9:57 PM Mike Ounsworth <Mike.Ou...@entrust.com> wrote:

Hi Sophie,

 

I’m not a great expert here, but I thought that X25519Kyber768Draft00 was an HPKE algorithm – therefore only applicable to the EncryptedClientHello, and not a general TLS KeyEx algorithm?


There are two different X25519Kyber768Draft00 — one for tls, and one for hpke:


That they are different is unfortunate, and hopefully will be rectified for ML-KEM. (Stay tuned.)

 

Genuine question: does the ECH need to be cryptographically self-contained, or is it ok for the security of ECH to depend on the TLS transcript check at the very end?


We don't want to have an HPKE KEM that is only safe to use within ECH. That'd be a footgun.

Best,

 Bas

Mike Ounsworth

unread,
Nov 27, 2023, 4:46:03 PM11/27/23
to Bas Westerbaan, Sophie Schmieg, bruno, pqc-...@list.nist.gov

 

---

Mike Ounsworth

 

From: 'Bas Westerbaan' via pqc-forum <pqc-...@list.nist.gov>
Sent: Monday, November 27, 2023 3:39 PM
To: Mike Ounsworth <Mike.Ou...@entrust.com>
Cc: Sophie Schmieg <ssch...@google.com>; bruno <bruno.p...@gmail.com>; pqc-...@list.nist.gov
Subject: Re: [EXTERNAL] Re: [pqc-forum] Fwd: Cloudflare public comments on FIPS IPD 203, 204, and 205.

 

On Mon, Nov 27, 2023 at 9: 57 PM Mike Ounsworth <Mike. Ounsworth@ entrust. com> wrote: Hi Sophie, I’m not a great expert here, but I thought that X25519Kyber768Draft00 was an HPKE algorithm – therefore only applicable to the EncryptedClientHello,

--

You received this message because you are subscribed to the Google Groups "pqc-forum" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pqc-forum+...@list.nist.gov.

Reply all
Reply to author
Forward
0 new messages