GENETIC ALGORITHM OPTIMIZATION OF NEURAL NETWORK HYPERPARAMETERS FOR PREDICTING KEY BITS IN THE S-AES CIPHER
Abstract
Recent advances in machine learning have opened new directions in the cryptanalysis of lightweight block ciphers, particularly in the study of nonlinear components and key-dependent transformations. Building on prior work involving simplified cryptographic models such as Mini-AES and deep-learning-based attacks on lightweight ciphers, this study investigates the learnability of round-key bits in the Simplified Advanced Encryption Standard (S-AES). A structured dataset was generated by producing random 16-bit master keys and deriving their corresponding 48-bit subkey representations through the key-schedule algorithm. Additionally, two fixed plaintext blocks were encrypted under each key to construct three distinct training sets for the classification of the KPK_PKP, KFK_FKF, and KSK_SKS round-key bits. To examine the predictive potential of machine-learning models, Support Vector Machines (SVMs) were chosen as primary classifiers due to their robustness and proven ability to capture nonlinear decision boundaries even in limited training regimes. The Ray Tune optimization framework was employed to identify optimal SVM hyperparameters, leveraging distributed search mechanisms that have demonstrated superior performance compared with conventional optimizers such as HyperOpt and SMAC.
References
[1] W. Zhang, E. Pasalic, “Highly nonlinear balanced S-boxes with good differential properties,” IEEE Transactions on Information Theory, vol. 60, no. 12, pp. 7970–7979, 2014.
[2] Yong Wang, Zhiqiang Zhang, Leo Yu Zhang, Jun Feng, Jerry Gao, Peng Lei, “A genetic algorithm for constructing bijective substitution boxes with high nonlinearity,” Information Sciences, vol. 523, pp. 152–166, 2020.
[3] H. Zahid et al., “Efficient Dynamic S-Box Generation Using Linear Trigonometric Transformation for Security Applications,” IEEE Access, vol. 9, pp. 98460–98475, 2021.
[4] M. Ahmad and M. Malik, “Design of chaotic neural network based method for cryptographic substitution box,” ICEEOT 2016, Chennai, India, pp. 864–868.
[5] R. C. Phan, “Mini Advanced Encryption Standard (Mini-AES): A Testbed for Cryptanalysis Students,” Cryptologia, vol. 26, no. 4, pp. 283–306, 2002.
[6] Kuryazov D.M, Sattarov A.B, Axmedov B.B, Blokli simmetrik shifrlash algoritmlari bardoshliligini zamonaviy kriptotahlil usullari bilan baholash, Tashkent, 2017.
[7] B. F. Abduraximov, J. R. Abdurazzoqov, “Deep learning-based cryptanalysis of Simplified AES,” Informatika va Energetika Muammolari, vol. 2, pp. 17–26, 2023.
[8] A. Bakhtiyor, B. Ilkhom, A. Javokhir, A. Orif, “Using the Capabilities of Artificial Neural Networks in the Cryptanalysis of Symmetric Lightweight Block Ciphers,” 2024, pp. 113–121.
[9] B. Y. Sun, D.-S. Huang, H.-T. Fang, “Lidar signal denoising using least-squares support vector machine,” IEEE Signal Processing Letters, vol. 12, pp. 101–104, 2005.
[10] A. Kazemi, R. Boostani, M. Odeh, M. R. Al-Mousa, “Two-Layer SVM: Towards Deep Statistical Learning,” EICEEAI 2022, IEEE, 2022.
[11] P. Chen, B. Wang, H.-S. Wong, D.-S. Huang, “Prediction of protein B-factors using multi-class bounded SVM,” Protein and Peptide Letters, vol. 14, no. 2, pp. 185–190, 2007.
[12] N. Cristianini, J. Shawe-Taylor, An Introduction to Support Vector Machines and Other Kernel-Based Learning Methods, 2000.
[13] C. J. Burges, “A tutorial on support vector machines for pattern recognition,” Data Mining and Knowledge Discovery, vol. 2, no. 2, pp. 121–167, 1998.
