Neuro-cryptanalysis of DES

In this paper, a new cryptanalytic attack on DES is presented. The proposed attack is a known-plaintext attack based on neural networks. In this attack we trained a neural network to retrieve parts of plaintext from ciphertext without retrieving the key used in encryption. The attack was practically, and successfully, implemented on DES. This attack required an average of 2¹¹ plaintext-ciphertext pairs to perform cryptanalysis of DES in an average duration of 51 minutes. As compared to other attacks, this attack is an improvement in terms of number of known-plaintexts required, as well as the time required to perform the complete attack.