Conference article

There are many (historical) unsolved ciphertexts from which we don’t know the type of cipher which was used to encrypt these. A ?rst step each cryptanalyst does is to try to identify their cipher types using different (statistical) methods. This can be difficult, since a multitude of cipher types exist. To help cryptanalysts, we developed a ?rst version of an artificial neural network that is right now able to differentiate between five classical ciphers: simple monoalphabetic substitution, Vigenère, Playfair, Hill, and transposition. The network is based on Google’s TensorFlow library as well as Keras. This paper presents the current progress in the research of using such networks for detecting the cipher type. We tried to classify all ciphers of a new MysteryTwister C3 challenge called “Cipher ID” created by Stamp in 2019. The network is able to classify about 90% of the ciphertexts of the challenge correctly. Furthermore, the paper presents the current state-of-the-art of cipher type detection. Finally, we present a method which shows that one can save about 54% computation time for classification of cipher types when using our artificial neural network instead of trying different solvers for all ciphertext messages of Stamp’s challenge.

cipher type detection; artificial neural networks; challenge; cryptanalysis; machine learning; TensorfFlow; Keras; Python