[TLS] Re: Concerns about the current draft.

[Tim Hollebeek <[e-mail]>]

Yes, this is not an AES thing. The fact that people pay so much attention to algorithms but often ignore modes is a problem, as I generally see more people using modes incorrectly than using the wrong algorithm.

 

In particular, X-ECB is horribly broken and these days probably should not be used by anyone, ever. That advice is already a decade old or more, at this point.

 

-Tim

 

From: Sophie Schmieg <sschmieg=[e-mail]> 
Sent: Friday, August 29, 2025 3:42 PM
To: John Mattsson <john.mattsson=[e-mail]>
Cc: [e-mail]
Subject: [TLS] Re: Concerns about the current draft.

 

Yes, technically you can parallelize Grover's you just lose the quadratic speedup, so it is not a very appealing tradeoff.

 

As for multiuser attacks, those only apply in a situation where the same plaintext is encrypted with a block cipher by multiple users. In other words, the attack only applies to deterministic modes such as AES-CMAC, AES-SIV or AES-ECB. And I do indeed not recommend using AES128-CMAC, AES128-SIV, or AES128-ECB (or AES256-ECB for that matter), in order to account for multiuser attacks. For the usage in TLS in particular, the IV is the xor of the sequence number and a per connection static IV, so multiuser attacks do not apply.

 

On Fri, Aug 29, 2025 at 11:17 AM John Mattsson <john.mattsson=[e-mail] <mailto:[e-mail]> > wrote:

Sophie Schmieg wrote:
> As Bas mentioned, there is currently no indication that Grover's algorithm can practically break AES-128 any time soon.
Completely agree.

>Grover's algorithm is inherently sequential, and cannot be parallelized,

Cannot be _effectively_ parallelized would be a more theoretically correct statement. But practically, Grover’s algorithm will never be used for breaking AES-128. Quantum computers will also remain much slower and much more expensive than classical computers. Anybody claiming Grover's algorithm is a threat to any cryptography should provide detailed calculation of the cost, size, and time. With realistic assumptions you end up with result like “a huge cluster of one billion CRQCs (according to one estimate costing one billion USD each) would take a million years of uninterrupted calculation to find a single AES-128 key” or “require qubits covering the surface area of the Moon”.

https://datatracker.ietf.org/liaison/1942/

https://www.youtube.com/watch?v=eB4po9Br1YY <https://www.youtube.com/watch?v=eB4po9Br1YY&t=3227s> &t=3227s 


John

From: D. J. Bernstein <[e-mail] <mailto:[e-mail]> >
Date: Friday, 29 August 2025 at 19:47
To: [e-mail] <mailto:[e-mail]>  <[e-mail] <mailto:[e-mail]> >
Subject: [TLS] Re: Concerns about the current draft.

Sophie Schmieg writes:
> Grover's algorithm is inherently sequential, and cannot be
> parallelized, making the standard approach of throwing more compute at
> the problem to scale up infeasible.

One can split the search space across many parallel quantum processors
and run a smaller Grover search on each part of the space, as noted in,
e.g., https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Farxiv.org%2Fabs%2Fquant-ph%2F0309123 <https://arxiv.org/abs/quant-ph/0309123> &data=05%7C02%7Cjohn.mattsson%40ericsson.com%7Ce91ec897b40744e464bc08dde72429dd%7C92e84cebfbfd47abbe52080c6b87953f%7C0%7C0%7C638920864700657440%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C&sdata=haVw1vVYALl5qAbq7VJPRciy2fAjPgrA0NEuAVbxyiY%3D&reserved=0 from Grover and Rudolph.

To some extent it's possible to combine this with multi-target attacks.
See https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fcr.yp.to%2Fpapers.html%23groverrho <https://cr.yp.to/papers.html#groverrho> &data=05%7C02%7Cjohn.mattsson%40ericsson.com%7Ce91ec897b40744e464bc08dde72429dd%7C92e84cebfbfd47abbe52080c6b87953f%7C0%7C0%7C638920864700717818%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C&sdata=CK8nBl6TfEM5LwY%2Fznl7KgJjT0x4dNiUG4IeSL7xf4s%3D&reserved=0.

Even without quantum computers, 2^40-target attacks against AES-128 are
already a feasible computation today for large-scale attackers. One can
try to stop these attacks by having every AES-128 application randomize
every input block, but we keep seeing people screw this up. This is the
sort of thing that _probably_ ends up working for TLS, but it's fragile,
and the broader habit of tolerating this sort of fragility is certainly
a big factor in the problems that we _have_ seen in TLS.

I agree that the Grover speedup compared to non-quantum searches comes
from the number of serial iterations carried out on each processor, and
meanwhile this has to fight against the quantum-computation overhead---
which could end up as 2^30 or 2^40; we don't know yet. But this doesn't
makes AES-128 a safe option: on the contrary, tolerating AES-128 will
end up compromising the confidentiality of some user data.

> confirmed the infeasibility of Grover's in at least the
> medium term of several decades to centuries

I see no basis for this claim.

> the reason for ECC + lattice based hybrids are becoming less
> compelling with every day that passes in which lattices do not get
> broken

One of the talks at Crypto 2025 last week said that none of the Kyber
parameters meet their claimed security levels.

---D. J. Bernstein

_______________________________________________
TLS mailing list -- [e-mail] <mailto:[e-mail]> 
To unsubscribe send an email to [e-mail] <mailto:[e-mail]> 

_______________________________________________
TLS mailing list -- [e-mail] <mailto:[e-mail]> 
To unsubscribe send an email to [e-mail] <mailto:[e-mail]> 




 

-- 


Sophie Schmieg | Information Security Engineer | ISE Crypto | [e-mail] <mailto:[e-mail]>