NIST 800-22 in Mathematica

The NIST Special Publication 800-22 is "A statistical test suite for random and pseduorandom number generators for cryptographic applications." Whew!

This provides a complete implementation of all the tests in Mathematica

The Implementation 07-Oct-06

Version 0.9.2 -- 07-Oct-06.

All code is licensed under the the standard BSD License

This is a preview release. Interfaces and names will change.

Specification Examples 07-Jul-05

Every example in the specification is recreated in order to verify the algorithms. A change log is included in the notebook.

Corrections 28-Jun-05

Unfortunately the specification contains a few errors. Here's what I found:

Test Status 28-Jun-05

Everything works, with some tests producing slight deviations from spec mostly due to numerical precision. The Spectral test is the lone oddball -- it "works" but the intermediate results do not match at all was is published. Use with caution and see notes below: 

1. Monobit Frequency Test -- OK
2. Block Frequency Test - OK
3. Runs Test -- OK, tau parameter not implemented
4. Longest Runs Test -- Ok, slight difference due to precision
5. Matrix -- Ok, possible numerical precision error in text.
6. Spectral -- Seems to work, but I can't get match the intermediate results.
7. Non-overlapping template -- OK -- G-SHA-1 test is not complete, but otherwise looks OK
8. Overlapping --OK, text has bad example
9. Mauers --OK, but need to implement main example
10. Lempel-Ziv -- OK, but I get an of-by-one count in main example
11. Linear Complexity -- ok
13. Approximate Entropy -- Ok
14. Cummulative Sums -- Ok, slight numerical deviatation in simple example.  Main example is same as specifiaction.
15. Random Excusions -- Ok,  pik table is exact, while specification uses 4 digits of precision causing slight discrepancy.
16. Random Excursions Variant-- OK, exact

TODO 28-Jun-05

not quite done yet...