ABSTRACT
Testing of the quality of random data is a very important issue, especially in cryptography. There are several ready to use test packages and recommendations. We have chosen FIPS140-2 tests that analyze 20,000 bit records and defi ne thresholds to assess TRNG randomness. Note that FIPS140-2 tests are really only a very basic statistical tests [1] and they cannot reveal all possible TRNG deviations. If more deep analysis is required it is possible to use, e.g., the NIST statistical test suite [17] that can analyze much longer bit records. Longer records are necessary for detection of very small statistical deviations. FIPS 140-2 consists of standard and easily understood statistical tests (if one of the tests fails, the generator fails the test) [18]:
Poker test Divide the 20,000 bit stream into 5,000 contiguous 4 bit segments. Count and store the number of occurrences of the 16 possible 4 bit values. Denote ( )f i as the number of each 4-bit value where 0 15i≤ ≤ . Evaluate the following:
Run test A run is defi ned as a maximal sequence of consecutive bits of either all ones or all zeros that is part of the 20,000 bit sample stream. The incidences of runs (for both consecutive zeros iG and consecutive ones iB ) of all lengths ( 1≥ ) in the sample stream should be counted and stored.
