Machine Learning in Cryptanalysis

Cryptographic algorithms are essential for securing digital data and communications. As computing technology advances and new attack methods emerge, it is crucial to continually evaluate these algorithms to ensure they remain secure against evolving threats. Traditionally, breaking cryptographic algorithms - known as cryptanalysis - has relied on mathematical and algorithmic techniques, requiring specialised expertise in cryptography.

In recent years, researchers have begun exploring data-driven approaches, using machine learning (ML) and deep learning (DL) models to assist in cryptanalysis.;These methods have shown promise, sometimes outperforming traditional attacks. However, they are often treated as "black boxes," meaning their decision-making processes are not well understood or interpretable. This project aims to advance the field of data-driven cryptanalysis by investigating;what ML and DL models learn about cryptographic systems. Using explainable AI tools such as SHAP (SHapley Additive exPlanations), we will analyse;how these models identify patterns or weaknesses in cryptographic algorithms.

By gaining a clearer understanding of their internal mechanisms, we aim to develop more effective and targeted attacks to evaluate the security of existing cryptographic systems. The outcomes of this research will contribute to strengthening digital security by developing advanced methods to test and identify vulnerabilities in cryptographic algorithms.

This work not only;enhances the field of cryptanalysis but also supports the development of more resilient cryptographic systems, ensuring the continued protection of;sensitive data in an increasingly interconnected digital landscape.