CRYPTOANALYSIS OF THE ADVANCED ENCRYPTION STANDARD (AES) FOR PLAINTEXT RESTORATION USING SEQUENTIAL NEURAL NETWORKS

Abstract

In this thesis, we investigate the potential of deep neural networks for cryptanalysis, with the aim of developing a network that can decipher encrypted data. Specifically, we use long short-term memory networks and a convolutional neural network to attempt to decrypt AES-encrypted plaint text and MNIST images, respectively. We find that while the network is unable to fully decipher encrypted data, it can successfully perform classification tasks on encrypted data. We also observe that the choice of encryption key has an impact on the network’s performance. Our findings suggest avenues for further research into the application of deep neural networks for cryptanalysis.