Re: [Hipsec] Need to close all draft-ietf-hip-dex-21 pending issues... before 2021-Jan-13...

[Eric Rescorla <[e-mail]>]

On Tue, Jan 19, 2021 at 7:29 AM Robert Moskowitz <[e-mail]>
wrote:

>
>
> On 1/19/21 9:44 AM, Eric Vyncke (evyncke) wrote:
>
> Bob,
>
>
>
> Thank you for your reply. I took some time to review the points. Look for
> EVY2> below but at first sight, this looks really good to me now. Of
> course, Roman and Ben will have a final say.
>
>
>
> About Ekr’s comment, I am reading your latest email sent minutes ago, and
> I am still unclear where, in the text, it is shown that FS is not
> achievable.
>
>
> 1.2 (and 1.2.1) cover it.  the cost of FS is beyond what 8-bit CPUs are
> reasonably able to handle.
>
> If some vendor is willing to pay the cost of BEX, it is always an option.
> Time and battery (and code) are against vendors going that route (choosing
> BEX for FS over DEX).
>

I must say, I find this whole discussion baffling.

To recap, the original rationale for this protocol was the one Bob makes
above, namely that "the cost of FS is beyond what 8-bit CPUs are reasonably
able to handle." However, this claim was presented without any actual
requirements for what an acceptable cost was and the protocol as sent by
the WGLC to the IESG included a wide range of cryptographic primitives,
some of which would be comparable if not slower to a forward secure
exchange with the best available algorithms (i.e., X25519)  This implies
one of three things:

1. The requirements are not known.
2. The requirements have quite a bit of headroom above a non-FS exchange
with the best available algorithms and potentially could accommodate FS.
3. The original protocol did not in fact meet the requirements.

The proper conclusion, in any case, is that we don't know whether we can
fit a FS exchange into the requirements and we won't until a proper
requirements analysis is done. Removing the NIST curves, as Bob proposes to
do, merely removes the obvious inconsistency from the specification; it
does not address the question of whether we need to abandon FS.

-Ekr




>
>
>
> Action plan: please apply the changes you have proposed below, post a
> revised I-D, then on my side:
>
>    - Launching an IETF Last Call for 2 weeks (as written below some
>    important changes have been made but I expect the LC eventless)
>    - Probing Ben & Roman to review whether their DISCUSS still hold
>
>
> I am ready with the changes.  I have an ICAO call I really need to get
> ready for, then a DR appt.  After that I will submit what I have if no new
> comments.
>
>
>    -
>
>
>
> Regards, it seems that the conclusion is near ;-)
>
>
>
> -éric
>
>
>
> *From: *Robert Moskowitz <[e-mail]> <[e-mail]>
> *Date: *Monday, 18 January 2021 at 17:06
> *To: *Eric Vyncke <[e-mail]> <[e-mail]>, Robert
> Moskowitz <[e-mail]> <[e-mail]>,
> "[e-mail]" <[e-mail]> <[e-mail]> <[e-mail]>,
> "[e-mail]" <[e-mail]>
> <[e-mail]> <[e-mail]>, Miika Komu
> <[e-mail]> <[e-mail]>
> *Cc: *Roman Danyliw <[e-mail]> <[e-mail]>, Eric Rescorla
> <[e-mail]> <[e-mail]>, Gonzalo Camarillo
> <[e-mail]> <[e-mail]>,
> "[e-mail]" <[e-mail]> <[e-mail]>
> <[e-mail]>, Benjamin Kaduk <[e-mail]> <[e-mail]>,
> Erik Kline <[e-mail]> <[e-mail]>, Adam Wiethuechter
> <[e-mail]> <[e-mail]>
> *Subject: *Re: [Hipsec] Need to close all draft-ietf-hip-dex-21 pending
> issues... before 2021-Jan-13...
>
>
>
>
>
> On 1/18/21 9:12 AM, Eric Vyncke (evyncke) wrote:
>
> TD ;LR : more work to be done, deadline this Thursday 21st
>
>
>
> Bob,
>
>
>
> Thank you for the -23 (and Adam W for the footwork) and I understand that
> you are quite busy.
>
>
>
> Here is the link to the diff between -21 and -23:
> https://www.ietf.org/rfcdiff?url2=draft-ietf-hip-dex-23&url1=draft-ietf-hip-dex-21 (i.e.,
> the one used by July 2020 IESG evaluation and the latest one)
>
>
>
> After the July 2020 IESG evaluation based on -21, there were a couple of
> points to be addressed (with some comments of mine as EVY>):
>
> 1.       Roman: “Section 6.3.  Per the definition of IKM, when should
> these two different derivations be used? "
>
> 1.       EVY> indeed, IKMm and IKMp are both defined but nothing is said
> which one to use in which case.
>
>
>        IKM       IKMm for Master Key SA Input keying material
>                  or
>                  IKMp for Pair-wise Key SA Input keying material
>
>        IKMm      Kij | I_NONCE
>        IKMp      Kij | I_NONCE | (concatenated random values of the
>                      ENCRYPTED_KEY parameters in the same order as
>                      the HITs with sort(HIT-I | HIT-R))
>
> Seems clear that IKMm is for the Master Key SA and IKMp is for the
> Pair-wise Key SA.
>
>
>
> EVY2> Suggest to add a pointer to section 2.3 (definitions) and possibly
> create entries for the master key SA and pair-wise key SA (perhaps a little
> extreme though)
>
>
>
>
>
>
> 2.       Roman "discuss-discuss" (read this as request for reply and
> non-blocking) about " further implementation experience provides better
> guidance" in sections 6 and 9.
>
> 1.       EVY> this really pleads for experimental status
>
>
> The only place this text exists anymore is in Appendix C:  iESG
> Considerations
>
> Perhaps I should delete it from there.
>
> EVY2> you can keep it the appendix, I guess this fixes Roman’s
> “discuss-discuss”
>
> 2.
>
> 3.       Benjamin on FOLD collisions
>
> 1.       EVY> IMHO addressed in the new section 3.2.1
>
>
> I believe I have this covered.  We have the Python scripts for tests, but
> this is a lot of code to put into the document.  Right now it is privately
> held by Adam and I.  If called on, we can find some permanent home for it.
>
>
> 2.
>
> 4.       Benjamin on ACL to counter FOLD collisions in section 3.2.1
>
> 1.       EVY> still light on the ACL but the above should clear it
>
>
> Sec 7.1 is referenced.
>
>
> 2.
>
> 5.       Benjamin "how is it known that the peer should be using DEX vs.
> BEX"
>
> 1.       EVY> partially addressed in section 1.2 but should be repeated
> in the security section
>
>
> I can create a sec 9.1 (pushing down the current 9.1):
>
> 9.1 Caution on using HIP DEX rather than HIP BEX
>
>    Due to the substantially reduced security guarantees of HIP DEX
>    compared to HIP BEX, HIP DEX MUST only be used when at least one of
>    the two endpoints is a class 0 or 1 constrained device defined in
>    Section 3 of [RFC7228]).  HIP DEX MUST NOT be used when both
>    endpoints are class 2 devices or unconstrained.
>
>
> Will this work?
>
>
>
> EVY2>  it works for me and hopefully for Ben as well
>
> 2.
>
> 6.       Benjamin lack of discussion on the security consequences of
> inadvertent counter reuse in AES-CTR
>
>
> See sec 9.1
>
> EVY2> Indeed, Ben’s ballot is dated March 2020 on the revision -13. It
> seems to be fixed to me in the latest revisions.
>
>
>
> 7.
>
> 8.       Benjamin "presence of a CSPRNG in order to obtain secure session
> keys"
>
>
> 9.  Security Considerations
>
> ....
>
>    *  The strength of the keys for both the Master and Pair-wise Key SAs
>       is based on the quality of the random keying material generated by
>       the Initiator and the Responder.  As either peer may be a sensor
>       or an actuator device, there is a natural concern about the
>       quality of its random number generator.  Thus at least a CSPRNG
>       SHOULD be used.
>
>
>
> EVY2> indeed, I should have detected that Ben was not using the -21 but
> the -13 😉 Now, please expand CSPRNG at first use.
>
> 9.
>
> 10.   Benjamin "usage of CMAC instead of HMAC" about KEYMAT algorithm
>
> 1.       EVY> new reference to NIST papers seems to address this concern
>
>
> Ben did agree in an email that the SP800-56C and 108 addressed the
> concern.  I did not need to go further.
>
>
>
> EVY2> if Ben was happy, then I am happy. Thank you for the information.
>
>
>
> 2.
>
> 11.   Ekr’s one about why forfeiting FS when some algorithm could do it
> in a reasonable time. In an email to authors and ADs, Eric R. wrote “it
> defines a set of parameters (the NIST curves) which are slower w/o FS than
> other parameters (X25519) are w/ FS. This fact calls into question the need
> to dispense with FS.”
>
> 1.       EVY> the additional section 1.2.1 and the reference to a paywall
> EfficientECC reference do not offer a conclusive motivation for an IETF
> standards w/o FS.
>
>
> Paywall?  Hmm.  I got it free.  I will have to check into this.  It may be
> to some cookie I have on this system.  Or the DOI has the wrong URL.
>
> Ah, that URL works for me because I am an IACR member.  For all else:
>
> https://link.springer.com/article/10.1007/s10623-015-0087-1
>
> So I will change the reference.  But please check this out.  I tried it on
> another machine that should not have my IACR cookies, but...
>
>
>
>
> 2.
>
>
>
> ***Bottom line, the document is not yet ready to be approved.*** (even if
> big progress has been made)
>
>
>
> As written in November (see below), the situation has lingered for too
> long and is blocking the HIP-NAT and rfc4423-bis documents.
>
>
>
> *** Therefore, I request the authors for a revised I-D addressing the
> above (and noting again that a change to ‘experimental’ – as there are no
> deployed implementations – could probably fix all of them) before Thursday
> 21st of January midnight UTC else I will ask the HIPSEC WG to agree
> removing the HIP-DEX section from the architecture document. ***
>
>
> Does the above address the open items?
>
>
>
>
> All in all, there have been a couple of significant changes (I_NONCE, some
> deleted ciphers) since the IETF last call (see
> https://www.ietf.org/rfcdiff?url2=draft-ietf-hip-dex-23&url1=draft-ietf-hip-dex-21
> ), so, another IETF Last Call is required but should not be a real problem.
>
>
>
>
>
> -éric
>
>
>
>
>
>
>
> From: Robert Moskowitz <[e-mail]>
> <[e-mail]>
>
> Date: Thursday, 14 January 2021 at 16:08
>
> To: Eric Vyncke <[e-mail]> <[e-mail]>, "Eric Vyncke
> (evyncke)" <evyncke=[e-mail]>
> <evyncke=[e-mail]>, "[e-mail]" <[e-mail]>
> <[e-mail]> <[e-mail]>, "[e-mail]"
> <[e-mail]> <[e-mail]>
> <[e-mail]>, Miika Komu <[e-mail]>
> <[e-mail]>
>
> Cc: Roman Danyliw <[e-mail]> <[e-mail]>, Eric Rescorla
> <[e-mail]> <[e-mail]>, Gonzalo Camarillo
> <[e-mail]> <[e-mail]>,
> "[e-mail]" <[e-mail]> <[e-mail]>
> <[e-mail]>, Benjamin Kaduk <[e-mail]> <[e-mail]>,
> Erik Kline <[e-mail]> <[e-mail]>
>
> Subject: Re: [Hipsec] Need to close all draft-ietf-hip-dex-21 pending
> issues... before 2021-Jan-13...
>
>
>
> I had hoped to get -23 out end of last week, and missed my cutoff.  I am
> now in IACR's Real World Crypto, where I have gotten a couple pointers for
> DRIP work.
>
>
>
> I was waiting for two analyzes that I got Jan 4, and incorporating them
> in.  I believe these SHOULD address much of EKR's questions.
>
>
>
> I will have a run of 1M DEX random HIs to HITs generated with no
> duplicates that I add in an Appendix along with the Python code.
>
>
>
> I am adding a BEX/DEX crypto cost into 1.2, probably 1.2.1:
>
>
>
> For an Initiator, BEX is:
>
>
>
> 2 PK sig varifications.
>
> 1 PK sig generation.
>
> 1 DH keypair generation.
>
> 1 DH secret derivation.
>
>
>
> DEX is:
>
>
>
> 1 DH secret derivation.
>
>
>
> I have cycles for these and a paper to reference, except ECDH keypair
> generation, on an 8 bit process and the numbers are big.  But I think that
> part belongs in an Appendix.
>
>
>
> So unlikely Friday.  But early the following week.
>
>
>
>
>
>
>
>
>
>
>
> On 1/12/21 6:19 AM, Eric Vyncke (evyncke) wrote:
>
> Two months after the email below, I sending a kind reminder to authors and
> WG.
>
>
>
> With the -22, a lot of (if not all ) SEC ADs’ DISCUSS points should have
> been addressed.
>
>
>
> As far as I can tell, the other remaining issue was Ekr’s one about why
> forfeiting FS when some algorithm could do it in a reasonable time. In an
> email to authors and ADs, Eric R. wrote “it defines a set of parameters
> (the NIST curves) which are slower w/o FS than other parameters (X25519)
> are w/ FS. This fact calls into question the need to dispense with FS.”
>
>
>
> While 2 months ago I put a deadline for tomorrow, I (as the responsible
> AD) am flexible of course but we cannot linger anymore. I know that a -23
> is in the work for weeks => let’s publish it in the coming days.
>
>
>
> Else, next week we will need to either change the intended status to
> experimental or declare the document dead by lack of energy. The latter
> does not have my preference obviously.
>
>
>
> Regards
>
>
>
> -éric
>
>
>
>
>
> From: Hipsec mailto:[e-mail] <[e-mail]> on
> behalf of "Eric Vyncke (evyncke)"
> mailto:evyncke=[e-mail]
> <evyncke=[e-mail]>
>
> Date: Friday, 13 November 2020 at 15:32
>
> To: mailto:[e-mail] <[e-mail]> mailto:[e-mail]
> <[e-mail]>, mailto:[e-mail]
> <[e-mail]> mailto:[e-mail]
> <[e-mail]>, Robert Moskowitz
> mailto:[e-mail] <[e-mail]>, Miika Komu
> mailto:[e-mail] <[e-mail]>
>
> Cc: Roman Danyliw mailto:[e-mail] <[e-mail]>, Eric Rescorla
> mailto:[e-mail] <[e-mail]>, Gonzalo Camarillo
> mailto:[e-mail] <[e-mail]>,
> mailto:[e-mail] <[e-mail]>
> mailto:[e-mail] <[e-mail]>, Benjamin Kaduk
> mailto:[e-mail] <[e-mail]>, Erik Kline mailto:[e-mail]
> <[e-mail]>
>
> Subject: [Hipsec] Need to close all draft-ietf-hip-dex-21 pending
> issues... before 2021-Jan-13...
>
>
>
> Dear HIP, dear authors,
>
>
>
> This document was requested for publication [1] in February 2018 (2.5
> years ago), then its IESG evaluation has been deferred, then I took over
> this document from Terry Manderson in March 2019, then it went again
> through IESG evaluation in July 2020 and there are still DISCUSS points to
> be addressed even after a couple of revised I-D...
>
>
>
> Difficult not to observe that this document does not progress very fast.
>
>
>
> Moreover, this document is a normative reference for rfc4423-bis waiting
> in the RFC editor queue since March 2019... So, also blocking the HIP-NAT
> document [2].
>
>
>
> After discussion with the HIP chair, Gonzalo in cc, we have taken the
> following decision: if a revised I-D addressing remaining DISCUSS points +
> Ekr’s ones is not uploaded within 2 months (13th of January 2021), then I
> will request the HIP WG to accept the complete removal of section A.3.3 of
> the rfc4423-bis document (1 page about HIP-DEX in the appendix) + the
> reference to the HIP-DEX document [3]. This will allow the immediate
> publication of the rfc4423-bis and HIP-NAT documents.
>
>
>
> The HIP DEX authors may also select to change the intended status of the
> document to ‘experimental’ (if the HIP WG agrees) as this may reduce the
> security requirements by the SEC AD and Ekr.
>
>
>
> Gonzalo and I are still hoping to get a revised HIP-DEX shortly,
>
>
>
> Regards
>
>
>
> -éric
>
>
>
> [1] https://datatracker.ietf.org/doc/draft-ietf-hip-dex/history/
>
> [2] https://www.rfc-editor.org/cluster_info.php?cid=C386
>
> [3] and possibly I will set the state of HIP-DEX as ‘dead’ on the
> datatracker
>
>
>
>
>
> --
>
> Robert Moskowitz
>
> Owner
>
> HTT Consulting
>
> C:      248-219-2059
>
> F:      248-968-2824
>
> E:      mailto:[e-mail] <[e-mail]>
>
>
>
> There's no limit to what can be accomplished if it doesn't matter who gets
> the credit
>
>
>
>
>