[149339] in cryptography@c2.net mail archive
Re: [Cryptography] cheap sources of entropy
daemon@ATHENA.MIT.EDU (James A. Donald)
Mon Feb 3 19:36:11 2014
X-Original-To: cryptography@metzdowd.com
Date: Tue, 04 Feb 2014 08:54:42 +1000
From: "James A. Donald" <Jamesd@echeque.com>
To: John Kelsey <crypto.jmk@gmail.com>, Jerry Leichter <leichter@lrw.com>
In-Reply-To: <51891D57-6166-49BE-BBC5-45B7858EE099@gmail.com>
Cc: Bill Stewart <bill.stewart@pobox.com>,
"cryptography@metzdowd.com" <cryptography@metzdowd.com>
Errors-To: cryptography-bounces+crypto.discuss=bloom-picayune.mit.edu@metzdowd.com
This is a multi-part message in MIME format.
--===============2069145336863445969==
Content-Type: multipart/alternative;
boundary="------------070403010209080008090105"
This is a multi-part message in MIME format.
--------------070403010209080008090105
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
On 2014-02-04 03:12, John Kelsey wrote:
> d. Code developed and even tested for one environment run on some new environment, and don't get any entropy.
We don't actually know this. By and large, the added complexity
provides added sources of random variation and unpredictability, rather
than suppressing existing random variation and complexity.
The case where turbulence induced timing variation would be lost is a
system that is fully cpu bound, and not IO bound. In such case, cache
hits and cache misses would depend on what all the other processes are
doing, which other processes are themselves dealing with things out
there, that have random variation, thus, random variation in cache hits
and cache misses, resulting timing variation dependent on all the real
external things that all the other processes have to deal with.
We can only measure turbulence randomness in a very simple, very
controlled system /because in a realistic system, there are a lot of
other sources of randomness/.
--------------070403010209080008090105
Content-Type: text/html; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
<html>
<head>
<meta content="text/html; charset=ISO-8859-1"
http-equiv="Content-Type">
</head>
<body bgcolor="#FFFFFF" text="#000000">
<div class="moz-cite-prefix">On 2014-02-04 03:12, John Kelsey wrote:<br>
</div>
<blockquote
cite="mid:51891D57-6166-49BE-BBC5-45B7858EE099@gmail.com"
type="cite">
<pre wrap="">d. Code developed and even tested for one environment run on some new environment, and don't get any entropy. </pre>
</blockquote>
We don't actually know this. By and large, the added complexity
provides added sources of random variation and unpredictability,
rather than suppressing existing random variation and complexity.<br>
<br>
The case where turbulence induced timing variation would be lost is
a system that is fully cpu bound, and not IO bound. In such case,
cache hits and cache misses would depend on what all the other
processes are doing, which other processes are themselves dealing
with things out there, that have random variation, thus, random
variation in cache hits and cache misses, resulting timing variation
dependent on all the real external things that all the other
processes have to deal with.<br>
<br>
We can only measure turbulence randomness in a very simple, very
controlled system <i>because in a realistic system, there are a lot
of other sources of randomness</i>.<br>
</body>
</html>
--------------070403010209080008090105--
--===============2069145336863445969==
Content-Type: text/plain; charset="us-ascii"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Disposition: inline
_______________________________________________
The cryptography mailing list
cryptography@metzdowd.com
http://www.metzdowd.com/mailman/listinfo/cryptography
--===============2069145336863445969==--
