An Efficiency Optimization Scheme for the On-the-fly Statistical Randomness Test

The randomness of random number generators (RNGs) significantly influences the security of cryptographic systems. Although RNGs are allowed to adopt in practical systems only after strict analysis and security evaluation, the randomness of generated sequences may degrade due to aging effects of electronic devices, change of temperature and humidity, or even malicious attacks. Therefore, before the generated sequence being used (as a secret key or any other critical cryptography parameter), it is necessary to execute the on-the-fly statistical randomness test (on-the-fly test) on the candidate sequence to ensure the security. On-the-fly test should be finished efficiently; otherwise, it would impact the cryptographic systems' performance. In this paper, we propose a scheme to optimize the efficiency of randomness test suites, that is, provide an optimized order of the tests in the test suite, so that an unqualified sequence can be rejected as early as possible. We apply this optimization scheme on the NIST test suite (SP 800-22) [1] as an instance. Experimental results of 128- and 256-bit sequence, demonstrate that the optimized efficiency approximates to the theoretical optimum and the scheme can be quickly implemented.