Pseudo-random sequence generation using the CNN universal machine with applications to cryptography

A good source of reproducible random-looking data is important in many applications ranging from simulation of physical systems, communications, and cryptography. It is demonstrated that the cellular neural network (CNN) universal machine (or the discrete-time CNN) is capable of producing a two-dimensional pseudo-random bit stream at high speeds by means of cellular automata (CA). First, the random properties of some irreversible two-dimensional CA rules, selected by applying mean-field theory, are analyzed by a battery of statistical tests. Second, a special class of reversible CA are considered for random number generation and are shown to have some of the desirable properties of physics-like models. Finally, as an example application for random number generation on the CNNUM, some cryptographic schemes are proposed.