Cryptoanalysis of Simple Substitution Ciphers with Genetic Algorithms

Abstract

Recent advances in the field of machine learning have once again raised the question whether computers can be trained to perform cryptanalytic tasks. In this paper, we identify the relationship between the machine learning and cryptanalysis with special attention on the genetic algorithms as a heuristic optimization method inspired by evolution processes in nature. We have indicated the consequences that new insights of machine learning may have on the reformulation of the practical criteria of secrecy in the synthesis of information security systems. Through the paper, we describe the approach based on genetic algorithms in order to confirm which machine learning algorithms are most suited for the purpose of the cryptanalysis and consequently to verify resistance of encryption algorithms to cryptanalysis.