UNBIASED RANDOM SEQUENCES FROM MEASUREMENTS OF POISSON PROCESSES

We consider the problem of true random bit generation from source vectors of independent geometric random variables, reduced modulo M for practical implementation. Independent geometric random variables result from measurements of discretized Poisson processes, which are good models for a number of physical sources. We propose a generalization of the classical approach by Elias, compute theoretical bounds, and evaluate the efficiency of the scheme by means of experiments. The proposed technique shows a significant advantage with respect to the classical approach.