An efficiency optimization scheme for the on-the-fly statistical randomness test

Abstract

In many cryptographic systems, random number can significantly influence its security. Although in practice random number generators (RNGs) are allowed to adopt only after strict analysis and security evaluation, the environmental factors also may lead the randomness of generated sequences to degrade. Therefore, on-the-fly statistical randomness test should be used to evaluate a candidate random sequence. Unfortunately, existing randomness test methods, such as the NIST test suite, are not well suitable to directly serve as on-the-fly test, because timely execution is usually not considered in their designs. In this paper, we propose a scheme to optimize the efficiency of randomness test suites, that is, providing the optimized order of the tests in a test suite, so that an unqualified sequence can be rejected as early as possible. This scheme finds out the optimized order by balancing the coverage, independence and time consumption of each test, and minimizing the average elimination time. We apply this optimization scheme on the revised NIST test suite as an instance. Experimental results on the sequences of 128 and 256 bits, demonstrate that the optimized efficiency approximates to the theoretical optimum and the scheme can be quickly implemented.