Without taking a position on the merits of these specific ideas,
I don't see why most of them would require a new version of
TLS. Indeed, a number of them (e.g., phasing out algorithms)
is something we routinely do without a version number bump.

it's possible that the HKDF one would, but even then I expect
we could simply have the definitions of how to use the new
hashes state that HKDF would not be used.

-Ekr


On Wed, Oct 1, 2025 at 11:35 PM John Mattsson <[email protected]>
wrote:

> >Are there any specific issues which cannot be addressed in the framwwork
> of TLS 1.3
>
>
>
> I think almost everything _can_ be handled with extensions. A new version
> is justified when changes become too large or too numerous to fit cleanly
> into the extension model. Another valid reason is to more clearly forbid
> outdated or insecure options.
>
>
>
> I don't think draft-zhou-tls-tls14 motivates 1.4. Traffic Flow Security
> (TFS) is already specified, and the other parts should likely not be done.
>
>
>
> That said, a TLS 1.4 could make sense sometime in the future if it were
> used to mandate stronger security properties and eliminate weak options.
> For example:
>
>
>
> - Require quantum-resistant cryptography only.
>
> - Require ephemeral key exchange in all cases.
>
> - Require implementations to perform mAEAD rekeying (currently left
> optional or often pushed to the application).
>
> - Phase out static keys, or at least not violate NIST requirements (SP
> 800-227).
>
> - Require support for strong rekeying. TLS 1.3’s KeyUpdate is much weaker
> than the mechanisms in IPsec, SSH, and Signal and does not provide
> protection against key leakage.
>
> - Provide identity protection. Today, PSK identities and SNI are still
> sent in the clear.
>
> - Avoid designing around HKDF and HMAC, which are only needed when the
> hash function has significant vulnerabilities/issues such as
> length-extension, failure to behave like a random function, lack of
> variable-length output, etc.
>
> - Mandate support for ECH and other extensions.
>
>
>
> Cheers,
>
> John
>
>
>
> *From: *Eric Rescorla <[email protected]>
> *Date: *Wednesday, 1 October 2025 at 19:59
> *To: *Bocai Zhou <[email protected]>
> *Cc: *[email protected] <[email protected]>, <[email protected]>
> *Subject: *[TLS] Re: [Pqc] Subject: Request for Technical Review:
> Internet-Draft draft-zhou-tls-tls14-03 – The Transport Layer Security (TLS)
> Protocol Version 1.4
>
> + TLS WG
>
>
>
> Bocai,
>
> I agree with Paul Hoffman that the place to send this
> document would be the TLS WG.
>
> With that said I don't believe that this document is necessary.
> It seems to make three primary technical changes:
>
> 1. Specifying a new set of PQ-only key establishment and signature
>    algorithms and new extensions to carry them.
> 2. Modifying CertificateVerify to support hybrid signatures and
>    Certificate to support two certificates.
> 3. The addition of the dummy packet for traffic analysis resistance.
> 4. The requirement for a specific PQ/traditional key establishment
>    combiner form.
>
> All of these can be accommodated within the existing framework
> of TLS 1.3.
>
> * For PQ-only key-establishment and signature, the client can simply
>   offer the existing code points with the existing extensions but not
>   offer any non-PQ values. There is no need to define new extensions,
>   and I think the semantics of your extension are worse, not better.
>
> * We already have an existing draft for dual certificate chains
>   within TLS 1.3 (
> https://datatracker.ietf.org/doc/draft-yusef-tls-pqt-dual-certs/).
>   Incidentally, I don't think your design is correct: you write:
>
>    If a server has negotiated a hybrid signature scheme, its Certificate
>    message MUST contain two certificates: one for a traditional
>    algorithm and one for a PQC algorithm.  These certificates MUST be
>    presented in a new certificate_list structure.
>
>   However, you need to present two *chains* not two certificates.
>   This is the reason for the structure in draft-yusef.
>
> * TLS 1.3 already supports empty application date records that just
>   consist of padding, as defined in S 5.4 of RFC 8446. Thus,
>   there is no need for a new record type to send cover traffic.
>
> * If we think your combiner form is better, we can just require it
>   in TLS 1.3.
>
> On the non-technical side, this document is largely the same as RFC
> 8446. Here in the IETF, we grant rights to our contributions to the
> IETF so there's nothing inherently wrong with just doing a -bis
> document like this, but it's also the case that most of the work
> in this document is really the work of other people, so it's not
> great to remove the Contributor list, thus denying them credit.
>
> -Ekr
>
>
>
>
> On Tue, Sep 30, 2025 at 9:22 PM Bocai Zhou <draft-ietf-tls-tls14=
> [email protected]> wrote:
>
> Dear esteemed PQUIP Working Group experts,
>
> I am writing to formally request a technical review of my Internet-Draft,
> draft-zhou-tls-tls14-03, which proposes The Transport Layer Security (TLS)
> Protocol Version 1.4.
>
> While the community’s current and critical post-quantum efforts are
> correctly focused on leveraging hybrid extensions within the established
> framework of TLS 1.3, this draft offers an alternative architectural
> exploration. I submit that the magnitude of the non-backward-compatible
> changes required for truly robust, long-term Post-Quantum Cryptography
> (PQC) integration—particularly concerning fundamental alterations to key
> derivation and authentication processes—warrants implementation under a new
> major protocol version number (TLS 1.4). This approach is designed to
> establish a cleaner, unambiguously secure, and sustainable foundation for
> PQC-era deployments.
> Core Technical Design Highlights for TLS 1.4
>
> The central divergence from the TLS 1.3 hybrid model is rooted in a
> dedicated separation of classical and PQC cryptographic components:
>
>    - Dedicated PQC Negotiation: TLS 1.4 introduces three new, dedicated
>    extensions: supported_pqc_groups, pqc_key_share, and
>    pqc_signature_algorithms. The existing classical supported_groups extension
>    is deliberately constrained to advertise only classical groups. This
>    cleanly enforces the separation of concerns for negotiation and simplifies
>    the implementation logic required for both hybrid and PQC-only modes.
>    - Key Combination Method: When a hybrid key exchange yields multiple
>    secret inputs (classical and PQC), TLS 1.4 strictly mandates the use of
>    length-prefixed concatenation to derive a single shared secret input
>    *before* its application to the HKDF-Extract function. This design
>    explicitly prohibits alternative mixing methods, such as XOR, thereby
>    enforcing cryptographic rigidity and soundness within the key schedule.
>    - Mandatory Hybrid Authentication: To effectively mitigate potential
>    downgrade and substitution attacks in the long term, the design requires
>    hybrid authentication to utilize two distinct certificate chains—one
>    classical and one PQC. Crucially, these chains must be cryptographically
>    linked (e.g., through cross-signatures or a Certified Linking X.509
>    Extension). The CertificateVerify message is accordingly updated to mandate
>    the inclusion and validation of both signatures over the identical
>    transcript hash.
>
> Specific Request for Expert Review
>
> Given that TLS 1.4 is fundamentally designed as a forward-looking,
> PQC-native protocol, the collective expertise of the PQUIP Working Group is
> invaluable to its validation. I respectfully request the working group to
> focus its review on the following critical areas:
>
>    1. The cryptographic soundness of the proposed key combination and
>    derivation method.
>    2. The security properties and operational feasibility of the
>    mandatory certificate linking mechanisms.
>    3. The clarity and robustness of the new negotiation rules and
>    extension separation.
>
> The current draft is available for your review here:
> https://datatracker.ietf.org/doc/draft-zhou-tls-tls14/03/
>
> I greatly appreciate your time and consideration of this proposal and look
> forward to a comprehensive discussion of this design on the mailing list.
>
> Best regards,
>
> Bocai Zhou
>
> Author, draft-zhou-tls-tls14
>
>
>
> --
> Pqc mailing list -- [email protected]
> To unsubscribe send an email to [email protected]
>
>
_______________________________________________
TLS mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to