Digital post-processing for testable random bit generators

In this paper, the problem of estimating the entropy produced by a post-processed random bit generator is discussed. A post-processing algorithm is proposed and a class of suitable sources is defined which includes stateless sources but also chaotic sources, provided that a state-reset function is implemented. It is shown that, using this general scheme of random bit generators, a straightforward procedure to evaluate the actual entropy delivered by a real device can be defined, thus supporting the device validation. The same procedure can be also used in simulation to evaluate the delivered entropy when the source is severely degraded. Finally, an attack scenario is defined and it is shown that it can be considered equivalent to a degraded source. The attack model allows choosing the post-processor compression ratio given the probability of forcing or probing each source output bit.