[Cryptography] "Anthropic A.I. Model Finds Flaws in Tough-to-Crack Encryption Algorithms"

[Tom Mitchell]

On Fri, Jul 31, 2026 at 1:01 PM Nicko van Someren <nicko at nicko.org> wrote:

> On Jul 31, 2026, at 02:17, Roland Ionas Bialke <
> webmaster at pro-wrestling.biz> wrote:
> ...
> > It sounds like the door opener device, where you have to enter a 4-digit
> code. At first it is 4^10 trys, but then - after a certain time of use -
> there is an abrasion on the buttons. So it is reducing the trys to 4^4.
>
>
One summary was it leveraged public hacked account data; twice.
Key management and account info just got more serious. Yes it already was.
Multi factor -- for sure.
It is tempting to use easy to type keys inside a net behind a firewall but
firewalls are hard to manage.
VPN from inside a home router especially with service provider hardware
seems to have more
risk than I would have considered a short time ago.  We often discuss the
mixed agenda of government
standards. Never forget to use plural in all these discussions.

Often a validation question was a pet name for password recovery.  Family
pet names are quite public.
Any tactic I discuss here becomes an attack surface.   Old curse... these
are interesting times.
Clusters all require shared secrets but many secrets should not be shared.
 Diplomatic codes were once
hand delivered.
--
Tom
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://www.metzdowd.com/pipermail/cryptography/attachments/20260731/fbc9691b/attachment.htm>