Machine Learning for RSA Cryptanalysis: Assessing Key Strength

RSA is one of the most used public-key cryptographic algorithms, and its security essentially depends on the size of the key and the proper generation of the key. As the power of computing grows and quantum threats emerge, identifying weaknesses in the RSA key structures becomes important to transition toward post-quantum cryptography. The paper presents a machine learning-based cryptanalysis approach to assess the strength of the RSA key and detect weak keys. Using a dataset of RSA key attributes, we extract features such as Modulus Bit Length, Private Exponent Bit Length, and Entropy to train various classification models, including Support Vector Machines (SVM), Random Forest, XGBoost, and Multi-Layer Perceptron (MLP). Our experimental results show that Modulus Bit Length is the dominant factor in determining RSA key strength, with models achieving near-perfect classification accuracy. It indicates that RSA weaknesses can be systematically exploited with techniques of machine learning, and thus transitioning to post- quantum cryptographic alternatives is urgent. This work was related to machine learning in cryptanalysis and gave insights on how cryptographic security can be assessed against rising computational threats.