Machine Learning Assisted Selection of Algebraic Structures for Post-Quantum Cryptographic Systems

Michael Nsikan John. (2026). Machine Learning Assisted Selection of Algebraic Structures for Post-Quantum Cryptographic Systems. Ktrend - International Journal of Computer Science and Artificial Intelligence (IJCSAI), 1(1), 1-16. https://doi.org/10.5281/zenodo.20651035

The migration from classical public-key cryptography to post-quantum cryptography requires systematic methods for identifying algebraic structures whose computational properties remain secure against both classical and quantum adversaries. Traditional selection is largely based on theoretical hardness assumptions, implementation cost, parameter constraints, and expert judgement. This paper develops a Machine Learning Assisted Algebraic Structure Selection Framework (MLASSF) for classifying candidate algebraic structures as suitable or unsuitable for post-quantum cryptographic use. The framework combines finite group invariants, conjugacy class statistics, character-theoretic indicators, elliptic-curve descriptors, lattice parameters, implementation efficiency, and estimated security categories into a supervised learning pipeline. A reproducible simulated dataset of 5,000 candidate structures was generated to model algebraic and computational features relevant to post quantum design. Five classifiers were trained and evaluated: logistic regression, support vector machine, random forest, gradient boosting, and artificial neural network. The best-performing model was gradient boosting, with an accuracy of 93.80%, precision of 95.04%, recall of 96.21%, F1-score of 95.62%, and ROCAUC of 95.29%. Random forest followed closely with 93.60% accuracy and also provided interpretable feature importance, showing that lattice dimension, implementation efficiency, character-degree variance, elliptic-curve rank, and derived length were influential predictors. The results show that machine learning can serve as a decision-support layer for algebraic selection in post-quantum cryptographic research, especially when combined with rigorous mathematical analysis rather than used as a substitute for proof-based security.