Advancing the RSA Cryptanalysis: An Experimental Demonstration of Plaintext Recovery Using Neural Networks
Vaideeshwaran Saravanan, Charlie Obimbo, Fatemeh Khoda Parast · Journal of Internet Technology and Secured Transaction · 2024
Cryptography is essential to cybersecurity since it guarantees data's secrecy, integrity, and validity against threats.The fast expansion of artificial intelligence (AI) in many applications needs the equal importance of mitigating AI-induced cybersecurity vulnerabilities and protecting AI from cyberattacks.This study investigates a plaintext recovery assault on the most common public key algorithm of RSA cryptographic technique using a 66-bit public key and a compact neural network with 1,276 parameters.The testing reaches Bit Probability Accuracy, precision, F1 score, recall, and specificity of an average of 95% for each, a system of combined bits of accuracy of 85% tested with a full dataset, and illustrating how AI may enhance cryptanalysis under certain situations by identifying three out of four plain texts from the amalgam of ciphertext and the public key.These results underscore the possible hazards linked to extensive AI systems that may train models with millions to billions of parameters, emphasizing the need for more research into AI's function in cryptanalysis.