Other Directory Sites: SeekWonder | Directory Owners Forum
The following archive was created by hippie-mail 7.98617-22 on Thu May 27 1999 - 23:44:21
David Honig (honig@sprynet.com)
Thu, 08 Apr 1999 17:55:25 -0700
At 05:17 PM 4/8/99 -0700, Jim Gillogly wrote:
>David Honig writes:
>> I just generated some raw binary files with my BSD 3.0 (Walnut Creek)
>> /dev/random and then ran Maurer's Universal Statistical Test (blocksize=8
>> bits)
>> on the result. MUST measured ~7.19 reliably on several indendent runs.
This
>> is the expected value for a uniformly distributed random sample.
>>
>> BSD's /dev/random is good.
>
>However (testing on Linux 2.2.0), /dev/urandom also gives ~7.19 over several
>runs. I also tested random(3) by printing out 8 bits of subsequent calls
>and running them through MUST, and they were also indistinguishable from
>/dev/urandom.
>
>Looks good, but MUST can't test everything.
Absolutely. Diehard looks for structure with multiple complex
(independent?) tests;
MUST gives you a single number. That's its utility and its flaw.
Diehard, however, requires more data (tens o' megabytes) than I was willing
to collect
at this time.
I will play with urandom and see if there's a BSD-diff.
NB: I've also used MUST to verify that IPsec is increasing the entropy of
my packets (using "tcpdump" to capture and the unix "yes" command to create
a stream of low-entropy chars). Of course, this cannot verify numerous
other vital
aspects of IPsec, but MUST is a fast, handy, lightweight pocket-knife for
the frontier cryptographer :-)
-------
Aside: The MUST metric uses compression-like operations to estimate the
negentropy of a sample of data. The metric is parametized by the size of
the blocks of bits taken as symbols. I used N=8 (bytes) for /dev/random.
-DH
The following archive was created by hippie-mail 7.98617-22 on Thu May 27 1999 - 23:44:21