[Seat] Re: Comments on formal analysis of relay attacks in intra-handshake attestation (CVE-2026-33697)

[camilo ayerbe <[e-mail]>]

I want to add a note from the applied side, as someone working on identity
and audit for agentic AI systems in regulated settings, where attested TLS
is increasingly load-bearing.

The finding in CVE-2026-33697 matters well beyond the formal-methods
discussion. For enterprises deploying confidential-computing workloads
under emerging regulatory obligations, a relay redirecting an attested
connection without breaking attestation is exactly the kind of gap that
undermines the attributable-and-auditable property these deployments are
meant to guarantee. The formal analysis making this concrete — and
reproducible — is a real service to practitioners, not just to the WG, and
it's the right yardstick against which the mitigations and alternative
mechanisms now being discussed should be measured.

I'll leave the model-level discussion to UFMRG and those closer to it. My
point is narrower: the class of problem this work surfaces matters for the
agentic-AI deployments many of us are building toward, and I'd encourage
the WG to treat rigorous attestation-binding analysis as core to that
effort.
Camilo Ayerbe
Ulissy s.r.l.


On Thu, Jul 9, 2026 at 12:08 AM Mark Novak <[e-mail]> wrote:

> Markus,
>
> Yes, we are submitting an I-D to IETF, and an early version was put in
> before the deadline, but it is not yet in a shape that I'm happy with --
> we'll be making further changes.
>
> Until then, please review the document on which that draft is based, it
> has most of the details.
>
>
> https://docs.google.com/document/d/1rWlKxbznOs1hycibPOTYDUxVSbXz7adOas-j5RFVA2I/edit?usp=drive_link
>
> You will need to request access, which I am happy to grant.
>
> On Wed, Jul 8, 2026 at 10:15 AM Markus Rudy <[e-mail]> wrote:
>
>> Hi Mark,
>>
>> Thanks for sharing details about your work! Did you already share a
>> proposal with the list? I would be interested in reading it, but can’t find
>> it right now.
>>
>> Cheers, Markus
>>
>> *From: *Mark Novak <[e-mail]>
>> *Date: *Wednesday, 8. July 2026 at 18:39
>> *To: *Steve <[e-mail]>; Songbo Bu <[e-mail]>
>> *Cc: *Muhammad Usama Sardar <[e-mail]>;
>> Nathanael Ritz <[e-mail]>; [e-mail] <[e-mail]>;
>> [e-mail] <[e-mail]>
>> *Subject: *[Seat] Re: Comments on formal analysis of relay attacks in
>> intra-handshake attestation (CVE-2026-33697)
>>
>> Humbly adding my 2c to this discussion.
>>
>> TLS protocol can remain unmodified if we separate the credential
>> acquisition flow from credential usage in TLS. The Trustworthy Workload
>> Identity SIG at the Confidential Computing Consortium has been working on
>> just such a proposal. We have defined several mechanisms by which a
>> workload, after performing Remote Attestation, can obtain a certificate to
>> use with TLS. This approach allows a client and a server end of a TLS
>> channel to be upgraded to Remote Attestation independent of each other,
>> which is an important factor in unlocking deployment.
>>
>> Our proposed mechanism does not suffer from the attacks mentioned in The
>> Register articles. We have performed extensive threat modeling to make sure
>> of that.
>>
>> I am looking forward to discussing this in Vienna with all interested
>> parties. Please engage with me in person or over email.
>>
>> ------------------------------
>> *From:* Steve <[e-mail]>
>> *Sent:* Monday, July 6, 2026 7:57 AM
>> *To:* Songbo Bu <[e-mail]>
>> *Cc:* Muhammad Usama Sardar <[e-mail]>;
>> Nathanael Ritz <[e-mail]>; [e-mail] <[e-mail]>;
>> [e-mail] <[e-mail]>
>> *Subject:* [Seat] Re: Comments on formal analysis of relay attacks in
>> intra-handshake attestation (CVE-2026-33697)
>>
>> Dear all,
>>
>> I am Research Analyst at Cloud Security Alliance (CSA-GCR).
>>
>> Because of highly severe attacks, the intra-handshake.fail paper deserves
>> close attention from the SEAT WG, UFMRG and the related research community.
>> >From a Cloud Security Alliance community and cloud-security industry
>> perspective, I strongly support bringing the paper and its supporting
>> open-source code into the WG/RG discussion.
>> Cloud and confidential-computing deployments rely on attestation
>> mechanisms that can be implemented, audited, and operated at scale. In this
>> setting, a protocol extension that adds intra-handshake attestation
>> behavior affects more than message syntax. It affects implementation
>> assurance, conformance testing, incident response, certification, and
>> enterprise risk acceptance.
>>
>> The paper is therefore important industry input because it connects a
>> concrete protocol-design question with formal analysis, relay-attack risk,
>> and production implementation impact. Its treatment of binding mechanisms
>> gives cloud providers, enterprise relying parties, security vendors, and
>> governance teams a sharper basis for judging whether a proposed attestation
>> transport path delivers security value proportionate to its operational
>> complexity.
>>
>> I strongly support using the Intra-handshake.fail paper and code as a
>> reference point for further SEAT and UFMRG discussion, and I encourage
>> review by WG participants, UFMRG experts, confidential-computing
>> implementers, and cloud-security stakeholders.
>>
>> Best regards,
>> Steve Luo
>> Research Analyst, Cloud Security Alliance (CSA-GCR)
>>
>> Songbo Bu <[e-mail]> 于2026年7月6日周一 22:18写道：
>>
>> Hi Usama, Slava, Jean-Marie, all,
>>
>> Based on my participation in the related research discussions, I
>> appreciate sharing Intra-handshake.fail paper and code with the WG/RG and
>> strongly support using them as important technical input for the ongoing
>> SEAT discussion.
>>
>> The paper and code is valuable because it makes the discussion more
>> property-driven. In particular, I think the WG should explicitly examine
>> what security property a hybrid intra-handshake plus post-handshake
>> attestation design can satisfy that post-handshake attestation alone cannot
>> satisfy. This question is independent of any particular specification
>> proposal, and it is important for evaluating whether additional
>> intra-handshake complexity is technically justified.
>>
>> I also support the paper’s narrower focus on intra-handshake attestation.
>> Keeping the analysis scoped in this way avoids conflating different
>> attestation models and allows the WG/RG to reason more precisely about
>> binding mechanisms, relay attacks, and the relationship between Evidence
>> and the TLS connection.
>>
>> >From WG charter perspective, the paper is useful because it gives the
>> community a concrete object to review, challenge, reproduce, and improve.
>> This is especially important where formal models, protocol state-machine
>> changes, and production implementations intersect.
>>
>> It helps bring the standards community and formal methods researchers
>> closer to share experience and ideas and provides UFMRG a concrete
>> methodology for protocol analysis of AI agents.
>>
>> For these reasons, I strongly support sharing this paper and code with
>> the WG/RG and encourage review from SEAT WG participants and UFMRG
>> formal-methods experts.
>>
>> Best regards,
>> Songbo Bu
>> _______________________________________________
>> Seat mailing list -- [e-mail]
>> To unsubscribe send an email to [e-mail]
>>
>> _______________________________________________
> Seat mailing list -- [e-mail]
> To unsubscribe send an email to [e-mail]
>