Cryptanalysis Attack on RSA using various Deep Learning Models
Sairaaj Surve, Shantanu Salvi, Ronnit Mirgh, Ramchandra Sharad Mangrulkar · 2024
Cryptanalysis systematically exploits vulnerabilities in cryptographic algorithms to decrypt ciphertext without access to the private key. This study leverages neural networks, which excel in identifying intricate patterns within large datasets, for cryptanalysis of RSA-encrypted text. Specifically, it investigates the use of various neural network architectures, including Deep Neural Networks (DNNs), Recurrent Neural Networks (RNNs), and Transformers, to generate decrypted text from encrypted inputs. The objective is to evaluate the effectiveness of these models in performing sequence-to-sequence tasks relevant to cryptanalysis. The experimental setup involves training and testing these models on a dataset generated by RSA encryption of an English language corpus using various key sizes. Results indicate varying levels of success across different architectures, with Transformers showing the best performance. This research contributes to understanding the applicability of neural networks in cryptanalysis and highlights their potential to identify vulnerabilities in cryptographic systems.