<!DOCTYPE html>
<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
</head>
<body>
<p><br>
</p>
<div class="moz-cite-prefix">Am 05.10.2026 um 04:20 schrieb Tom
Mitchell:<br>
</div>
<blockquote type="cite"
cite="mid:CAAMy4URPWcJ7uMMx4ZyT-QTewbt4-h8TjG421X30tpRdST4uvQ@mail.gmail.com">
<meta http-equiv="content-type" content="text/html; charset=UTF-8">
<div><br>
</div>
<div><br>
<div class="gmail_quote gmail_quote_container">
<div dir="ltr" class="gmail_attr">On Sun, Oct 4, 2026 at 17:18
zeb--- via cryptography <<a
href="mailto:cryptography@metzdowd.com"
moz-do-not-send="true" class="moz-txt-link-freetext">cryptography@metzdowd.com</a>>
wrote:<br>
</div>
<blockquote class="gmail_quote"
style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"
dir="auto">> From: Jens Weber<br>
> Sent: Thursday, 1 October 2026 23:35<br>
> To: <a href="mailto:cryptography@metzdowd.com"
target="_blank" moz-do-not-send="true"
class="moz-txt-link-freetext">cryptography@metzdowd.com</a><br>
> Subject: Re: [Cryptography] Cryptographic feasibility
of a decentralized trust infrastructure without a master key<br>
> <br>
> Ideally, AI functions would operate only within
narrowly authorized contexts and should not require a global
view of all sensitive data.<br>
<br>
I think Data would disagree. I do too.</blockquote>
<div dir="auto"><br>
Data as in Star Trek 😳</div>
<div dir="auto"><br>
</div>
<div dir="auto">Large systems, disagreement is expected.</div>
<div dir="auto">So build a Ven-diagram of owners, consumers,
customers, users AND abusers of the data and system.</div>
<div dir="auto">A second diagrammatic view of applicable laws
including a globe of governments AND note that laws are
dynamic. So any partitioning and mandatory access lists
must be ephemeral. Even necessary audit data will have
bounds.</div>
<div dir="auto"><br>
</div>
<div dir="auto">AI functions are interesting, map mandatory
access controls into AI systems much as one might for an SQL
data system. The Rainbow Series for trusted systems
evaluation is a start. No application will be considered
secure if the platform (model) is not secure. </div>
<div dir="auto"><br>
</div>
<div dir="auto">Addition of features needs to be considered.
In high school math I was taught the value of factorial
computing. I was also taught about orders of infinity. The
difference between two giant (infinite) sets can be small —
see Aleph-null vs. Aleph-1</div>
<div dir="auto">The implied combinations of “stuff” gets large
quickly. <=== this</div>
</div>
</div>
<br>
<fieldset class="moz-mime-attachment-header"></fieldset>
<pre wrap="" class="moz-quote-pre">_______________________________________________
The cryptography mailing list
<a class="moz-txt-link-abbreviated" href="mailto:cryptography@metzdowd.com">cryptography@metzdowd.com</a>
<a class="moz-txt-link-freetext" href="https://www.metzdowd.com/mailman/listinfo/cryptography">https://www.metzdowd.com/mailman/listinfo/cryptography</a>
<span style="white-space: normal;">Tom,</span></pre>
<p><span>thank you again for your criticism and suggestions. Your
comments about dynamic laws, bounded audit data, access
controls and the growing complexity of large systems have been
very useful.</span></p>
<p><span>I have continued modelling NOX without selecting a
blockchain or a specific cryptographic implementation.</span></p>
<p><span>One problem now appears particularly difficult:</span></p>
<p><span>Assume a hospital temporarily loses network connectivity.
A physician continues working with locally available
credentials. During that outage, the physician's credential is
revoked elsewhere because the key has been compromised.</span></p>
<p><span>When the hospital reconnects, the system has to
determine:</span></p>
<ul>
<li><span>what the hospital knew at the time,</span></li>
<li><span>whether the credential was valid according to its last
known state,</span></li>
<li><span>when the revocation actually became effective,</span></li>
<li><span>whether offline actions should remain valid, be
flagged for review, or be rejected,</span></li>
<li><span>and how the ordering of these events can be
demonstrated without trusting only the local device clock.</span></li>
</ul>
<p><span>I do not want an offline mode to become a disguised
master key or a general bypass of authorization.</span></p>
<p><span>This leads me to two questions for Tom and anyone else on
the list who would like to comment:</span></p>
<ol start="1">
<li>
<p><span>Are short-lived credentials combined with later
reconciliation a reasonable architectural direction for
such a system?</span></p>
</li>
<li>
<p><span>How would you approach trustworthy event ordering and
revocation during network partitions without depending on
one global trusted clock?</span></p>
</li>
</ol>
<p><span>I am deliberately leaving the cryptographic mechanism
open at this stage. I am mainly interested in whether the
trust model itself is sensible or whether there is a
fundamental flaw in the assumptions.</span></p>
<p><span>Best regards,</span><br>
<span>F. Ronecker</span></p>
<pre wrap="" class="moz-quote-pre">
</pre>
</blockquote>
</body>
</html>