[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Noise sphere graphical tests of randomness





I've got a short program for the PC (w/TPascal source) that plots a 
noise sphere from a file of (pseudo) random data, if anyone is 
interested.  Requires a VGA card that handles mode 5Fh (640 x 480, 
256 color) though the source can be eaily recompiled to use something 
else, or standard Borland Graphics drivers.

Send a reply with the subject "send nsphere" or check  ftp.funet.fi 
in the /pub/crypt/random directory in a few days [the ftp site would be
easier on my mailer ;]

The source has a brief explanation of what noise spheres are and a 
reference to the Pickover article the program was based on.

The source code is not copyrighted, and it would be nice to see it ported to
other systems, or maybe a portable C version that writes the output to a
.PCX file in RBG(?). [I'm not that fluent in C to write one...]

I've gotten some *very* interesting results.  Raw samples collected 
from fast timings between Windows message broadcasts do nicely with 
some of the randomness tests (compressability, Maurer, chi-sq) but 
clearly plots a spiral. 

Raw samples from the keyboard don't do as nicely in other tests (~50%
compressability, <6 bits/byte Maurer, and "non-random" in chi-squared] but
in the plot no discernable pattern shows up.... similar to plotting the
output from /dev/urandom [even when /dev/urandom was configured *not*
to use keyboard...]

Needless to say this will affect sampling methoids in the next version of
NOISE.SYS.

take care,

  Rob

---
Send a blank message with the subject "send pgp-key" (not in
quotes) to <[email protected]> for a copy of my PGP key.