Cryptographic Algorithm Identification Using Deep Learning Techniques

摘要

Deep learning techniques have recently gained momentum in cryptography and cryptanalysis. Deep learning architectures such as deep neural networks, deep belief networks, recurrent neural networks and convolutional neural networks have diverse applications in diverse fields. Encryption algorithm identification by analysis of cipher text enables focused cryptanalysis methods to be applied, thereby increasing the chances of a successful full/partial plain text recovery. In this work, we identified the encryption algorithm used, by analyzing various encrypted text files. We applied modern deep learning classification methods and compared them with existing machine learning methods. We generated cipher text corpus for multilingual dataset that contains 700 files with an average of 4000 characters per file and applied our techniques on this generated ciphertext. We used ciphers namely AES and Blowfish.