Statistical Test for Randomness.
Venkata Koonaparaju
Specification initial submission
ECE 746
Introduction, Motivation and Originality
Most of the practical random number generators have some built in regularities which, sometimes, also
become their deficiencies. Usage of these generators in cryptographic applications will need more
security requirements than other applications. There are mathematical safety measures that offer some
protection against this issue. One safety measure is to assess generators by theoretical tests and the
other is to run empirical tests. Since the year 2000 NIST has been developing a string of statistical test to
predict the efficiently predict the randomness of a string of bits. A total of 16 tests were developed,
implemented and tested as of today. All of these tests were implemented in C by NIST. The purpose of
this project is to implement at least 8 of the tests in C#. This project can be extended to implement all
the 16 of them and can also be used in CrypTool 2.0 in future.
Language and Platform
This project will use .net framework 3.0 and will run on Microsoft windows platform. Visual Studio 2008
will be used to take advantage of new technologies like WPF and C# is used as the base language
because of its popularity and modern concepts and design. Even though application developed in .net
run mostly on windows platform there are applications out in the market like Mono 1.0 which will let
these applications run on other platforms like Sun Solaris. This project doesn’t look into the cross
platform implementations and will use windows platforms.
Additional Software Required
This project uses the C code developed by NIST for reference purposes and also might use them for
testing purposes.
Testing Strategy
The following steps will be involved in testing the implemented algorithms
1) Using a standard random number generator a set n bit sequences will be generated and stored
in a text file
2) The tests implemented in C# will be run on the text file
3) A output file will be generated with relevant intermediate values and like P Values based on
which conclusions can be drawn
