Applying hierarchical clustering to homophonic substitution ciphers using historical corpora

Abstract This study examines whether hierarchical clustering can be used efficiently on homophonic substitution ciphers from the early modern age. The methodology was first tested on artificial homophonic ciphers based on historical corpora in 4 different languages. This investigation established that the decryptability of homophonic substitution letters does not depend on the plain text language, and it is effective if the text length goes beyond the unicity point. Finally the examination of original homophonic substitution cipher – cipher-key pairs followed in 4 languages. It determined that the method is applicable to original historical encrypted documents as well.