Evolving Cryptographic Pseudorandom Number Generators

Random number generators (RNGs) play an important role in many real-world applications. Besides true hardware RNGs, one important class are deterministic random number generators. Such generators do not possess the unpredictability of true RNGs, but still have a widespread usage. For a deterministic RNG to be used in cryptography, it needs to fulfill a number of conditions related to the speed, the security, and the ease of implementation. In this paper, we investigate how to evolve deterministic RNGs with Cartesian Genetic Programming. Our results show that such evolved generators easily pass all randomness tests and are extremely fast/small in hardware.