On a classification of finite statistical tests

Statistical tests of random sequences are often used in cryptography in order to perform some routine checks for random and pseudo-random number generators. Most of the test suites available are based on the theory of hypothesis testing which allows one to decide whether a sample has been drawn following a certain distribution. In this article, we develop a theoretical foundation of statistical tests of random sequences and hypothesis testing with a focus on cryptographic applications and we draw some interesting practical consequences.