Cryptography is fundamental to cybersecurity, as it ensures confidentiality, integrity, and authenticity of data against adversaries. With the rapid growth of artificial intelligence (AI) across diverse applications and systems, addressing AI-driven cybersecurity threats is as important as safeguarding AI from cyberattacks. This research explores a plaintext recovery attack on the widely used RSA cryptographic algorithm with a 66-bit public key, utilizing a small neural network with 1,276 parameters. The attack achieves a Bit Probability Accuracy of 95%, demonstrating how AI can assist in optimizing cryptanalysis under certain conditions. This suggests potential vulnerabilities in RSA when combined with advanced AI techniques, raising concerns about future security challenges, especially with quantum algorithms such as Shor’s algorithm. Furthermore, the findings emphasize the potential risks associated with large-scale AI systems capable of training models with millions to billions of parameters, highlighting the need for further investigation into AI’s role in cryptanalysis.

Authors: Vaideeshwaran Saravanan, Charlie Obimbo, Fatemeh Khoda Parast

Published in: International Conference for Internet Technology and Secured Transactions (ICITST-2024)

• Date of Conference: 4-6 November 2024
• DOI: 10.20533/ICITST.2024.0017
• ISBN: 978-1-913572-76-1
• Conference Location: St Anne’s College, Oxford University, UK