[Rats] Re: Threat model and properties for remote attestation

[Manu Fontaine <[e-mail]>]

On Thu, Aug 13, 2026, 4:23 AM Thomas Fossati <[e-mail]> wrote:

What confidential computing can achieve (if properly deployed) is...
>

Thomas, you tie what Confidential Computing capabilities to conventional
deployment assumptions. When dealing with a technology that inherently
shrinks the attack and trust surfaces of a hardware+software actor by
orders of magnitude, the North Star should be to explore, design, and
standardize a system around it that minimizes the degradation of that
initial security posture, i.e. the entire system should be built the same
way. From that idealistic blueprint, you can decide where current reality
forces compromises.

If RATS WG took such a complete system design perspective, it would see
that there is a path that minimizes trust dependencies beyond what you
state is possible. We know because we do operate a recursive and
hierarchical network of "verifiers of verifiers", entirely built on CC,
that does not have some of the trust dependencies you implicitly assume
cannot be removed.

I know that's not how the RATS WG works, and unfortunately it may lead some
folks to conclude that CC has limitations that are, in fact, imposed by
legacy deployment approaches, not the technology itself.

Thanks,
M