Modelling cryptographic distinguishers using machine learning

Abstract Cryptanalysis is the development and study of attacks against cryptographic primitives protocols. Many properties rely on difficulty generating an adversary who, given object sampled from one two classes, correctly distinguishes class used to generate that object. In case cipher suite distinguishing problem, classes are different primitives. this paper, we propose a methodology based machine learning automatically classifiers can be by solve any problem. We discuss assumptions, basic approach for improving advantage as well phenomenon call “blind spot paradox” . apply our distinguishers NIST (DRBG) Finally, provide empirical evidence might statistically have some distinguish between DRBG used.