Perfect Secrecy under Deep Random assumption

We present a new idea to design perfectly secure information exchangeprotocol, based on a new form of randomness, called « Deep Randomness »,generated in such a way that probability distribution of the output signal ismade unknowledgeable for an observer. By limiting, thanks to DeepRandomness, the capacity of the opponent observer to perform bayesianinference over public information to estimate private information, we candesign protocols, beyond Shannon limit, enabling two legitimate partners,sharing originally no common private information, to exchange secretinformation with accuracy as close as desired from perfection, andknowledge as close as desired from zero by any unlimitedly poweredopponent. We introduce the Deep Randomness assumption, based on PriorProbability theory developed by Carnap, Cox and Jaynes. Deep Randomassumption gives a rigorous rule to perform bayesian inference underuncertainty situation. We also present a workable example of protocol withproof of perfect secrecy under Deep Random assumption, and we present amethod to generate Deep Randomness from classical computation resources.