Spectra is a simple tool designed to aid in evaluating the quality of random numbers generated by TRNGs and PRNGs. It does this by counting the occurrences of characters within a given input file to help detect perceived bias, and then visualizes the data as colored (or black and white) pixels in an image file to aid in detection of repeating patterns.
As the human mind is exceptionally good at picking up on patterns that might otherwise be difficult to detect mathematically, Spectra allows the user to rapidly make a determination on the repeatability of numbers generated from the suspect RNG.
Spectra requires relatively recent versions of GD and libpng. Spectra does not require X to run, though you will obviously want a graphical environment when trying to examine large images.
To compile Spectra, simply run the command "make" in this directory. If you get any message about missing libraries, use your distribution's package manager to install them. This will let you run Spectra from the current directory for a quick test, but not install it to the system.
To actually install Spectra, use the command "make install". Since this will install Spectra to the system directories, you will need to run this as root or through sudo.
When and if you want to remove Spectra, you would use the command "make uninstall". As with "install", this makes changes to the system directories and will therefore require root permissions.
In addition, if you are upgrading from a previous version, there is also the command "make upgrade". This will remove any older versions of Spectra, compile the new version, and finally install it to the system. As you might have guessed, this also requires root permissions.
Spectra takes an ASCII input file consisting of digits between 0 and 9. Any other character in the file will cause Spectra to abort. Currently, Spectra only supports output to PNG files, as JPG compression is too lossy and other file formats (like BMP) are prohibitively large.
As of the current version, Spectra supports the following options:
-i
-x
-y
Once it has finished rendering the image, Spectra will give you a breakdown of how many times each character appeared in the file. This will give you a good idea of the data's bias. Too many of one number is not a good sign, the spread should be as close to even as possible.
Looking at the actual image file itself will allow you to visually detect any patterns in the generated data. For example, a file that contained alternating 0's and 1's would show equal occurrences (which would be good), but visually would appear as a mesh or grid type design.
An image from a good data source (Random.org) and an image from a very bad data source are included under the /examples directory that came with the source. While most RNGs won't look nearly as bad as the example given here, it should give you an idea as to what to look for. Basically, any repeating pattern or obvious clustering of pixels is a bad sign.
Currently, Spectra uses a static color-to-character mapping for the rendered image. The colors are as follows:
0 - Black 1 - White 2 - Red 3 - Orange 4 - Yellow 5 - Green 6 - Blue 7 - Aqua 8 - Pink 9 - Purple
Spectra's Makefile supports using environment variables to modify the installation so it can be more easily built into a binary package for your distribution. You can pass compiler options and the installation directory to the Makefile like so:
$ make CFLAGS="-O2 -march=i486 -mtune=i686" gcc -O2 -march=i486 -mtune=i686 -lgd -lpng -lm spectra.c -o spectra $ mkdir ./pkg $ make install DESTDIR=./pkg mkdir -p ./pkg/usr/bin/ mkdir -p ./pkg/usr/share/doc/spectra-1.2/ cp spectra ./pkg/usr/bin/ cp -a ChangeLog COPYING README examples/ ./pkg/usr/share/doc/spectra-1.2/
This allows you to easily script the build process and create a package using the package creation software for your distribution of choice.
If you are able to build packages for your distribution, please let me know and we can work something out for hosting them on the site.
This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License version 2 as published by the Free Software Foundation.
This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
For details, see the file "COPYING" in the source directory.
For more information and updates, please see www.digifail.com