Machine learning assisted cryptanalysis of block ciphers
*Anusha SreeramCorresponding authorseeramanusha@gmail.comFaculty of Operations and Information TechnologyICFAI Business SchoolThe ICFAI Foundation for Higher Education (IFHE) (Deemed to be University)Hyderabad, Telangana, 501203, IndiaView full profile → , Priya Chaudharypriyachaudhary@outllook.comDepartment of Computer Engineering & ApplicationsGLA UniversityMathura, Uttar Pradesh, 281406, IndiaView full profile → , Rajeev Sharmarajeev.sharma.orp@chitkara.edu.inCentre for Research Impact and OutcomeChitkara UniversityRajpura, Punjab, 140417, IndiaView full profile → , Prakhar Goyalprakhar.qsb@quantumeducation.inQuantum University Research CenterQuantum UniversityRoorkee, Uttarakhand, 247667, IndiaView full profile → , Kaustubh Milind Utpatkaustubh.utpat@vupune.ac.inDepartment of Mechanical EngineeringVishwakarma UniversityPune, Maharashtra, 411048, IndiaView full profile → , Vudathu Sobhasobhavudathu04@gmail.comDepartment of Computer ScienceAditya Degree & PG Colleges (Autonomous)Kakinada, Andhra Pradesh, 533003, IndiaView full profile → , Yagna B. Adhyaruybadhyaru.ce.hcet@gokuluniversity.ac.inFaculty of EngineeringGokul Global UniversitySidhpur, Gujarat, 384151, IndiaView full profile →
* Corresponding author · click or hover a name for details
- Received:
- 01 Apr 2026
- Published Online:
- 31 Oct 2026
- Article type:
- Research Article
- Language:
- EN
- Article no.:
- JDMSC-3073
- Pages:
- 4557–4570
Abstract
The growing use of secure digital communication systems in various fields like cloud computing, Internet of Things (IoT), and critical infrastructure has heightened concerns on the need to have strong cryptographic mechanisms. Block ciphers continue to be a stalwart of the data security of the modern era but their increasingly complex nature and inability to be attacked using conventional analytical methods pose a major challenge to successful cryptanalysis. Traditional methods like linear and differential cryptanalysis may need a lot of computing power and sizeable datasets, restraining their use to current lightweight and nonlinear encryptions. This brings an urgent demand of adaptive, scalable and intelligent cryptanalysis techniques. In this paper, a cryptanalysis framework of block ciphers with the help of machine learning is suggested, which would improve the effectiveness of key recovery by using data-oriented methods. The main goal of the research is to fill the gap between the classical cryptanalysis and the current application of artificial intelligence in the form of supervised learning models, to detect latent patterns in ciphertext transformations and to decrease the effective key search space. The proposed framework incorporates the structured dataset creation, feature engineering based on bit-level and differential representations and implementation of machine learning algorithms, such as Deep Neural Networks (DNN), Convolutional Neural Networks (CNN), and Support Vector Machine (SVM). The approach includes the creation of plaintext ciphertexts of various block ciphers, preprocessing the data, and extracting meaningful features, and training predictive models to predict key-related outputs. Standard metrics are used to assess the performance of the framework, including accuracy, precision, recall, F1-score, and Key Recovery Rate (KRR), and computational metrics, including training time and inference efficiency. The experimental findings prove the suggested method to be much more successful than conventional cryptanalysis methods and estimates the accuracy level to 96.1 percent and KRR to 93 percent, as well as reduces the computational complexity compared to exhaustive search methods.
Keywords
Subject Classifications
References
[1] A. Al-Sabaawi, “Cryptanalysis of classic ciphers: Methods implementation survey,” in Proc. Int. Conf. Intelligent Technologies (CONIT), pp. 1–6 (2021).
[2] A. K. Al Hwaitat, M. A. Almaiah, A. Ali, S. Al-Otaibi, R. Shishakly, A. Lutfi, and M. Alrawad, “A new blockchain-based authentication framework for secure IoT networks,” Electronics, vol. 12, no. 17, Art. no. 3618 (2023), doi: 10.3390/electronics12173618.
[3] S. O. Akindeinde and S. D. Yakubu, “Hybrid block method with quasilinearization for the dynamical simulation of a methamphetamine abuse model,” J. Interdiscip. Math., vol. 29, no. 2, pp. 399–412 (2026), doi: 10.47974/JIM-2337.
[4] A. Kariya and B. K. Behera, “Investigation of quantum support vector machine for classification in NISQ era,” arXiv preprint arXiv:2112.06912 (2021).
[5] A. Tiwari, S. C. Addimulam, N. Dubey, K. K. Kumar, D. Kapila, and K. Jain, “Cybersecurity in voice over IP (VoIP) systems: Protecting communication networks from threats, vulnerabilities, and cyber attacks,” J. Discrete Math. Sci. Cryptogr., vol. 28, no. 8, pp. 2989–2999 (2025), doi: 10.47974/JDMSC-2443.
[6] R. Pakrooh, A. Jabbari, and C. Fung, “Deep learning-assisted security and privacy provisioning in the Internet of Medical Things systems: A survey on recent advances,” IEEE Access, vol. 12, pp. 40610–40621 (2024).
[7] C. Shao, D. Zhao, H. Li, S. Cheng, S. Gao, and L. Yang, “Detection of security vulnerabilities in cryptographic ICs against fault injection attacks based on compressed sensing and basis pursuit,” J. Cryptogr. Eng., pp. 1–14 (2023).
[8] S. Sikdar and M. Kule, “Recent trends in cryptanalysis techniques: A review,” in Proc. Int. Conf. Frontiers in Computing and Systems, pp. 209–222 (2022).
[9] J. So, “Deep learning-based cryptanalysis of lightweight block ciphers,” Secur. Commun. Netw., vol. 2020, Art. no. 3701067 (2020).
[10] W. Villegas-Ch, R. Gutierrez, I. Sánchez-Salazar, and A. Mera-Navarrete, “Adaptive security framework for the Internet of Things: Improving threat detection and energy optimization in distributed environments,” IEEE Access, vol. 12, pp. 157924–157944 (2024).
[11] S. Simhadri, Raghavendra, and B. R. Purushothama, “AI-Powered Cryptanalysis: Identifying Encryption Algorithms and Recovering Plaintext,” 2025 International Conference on Networks and Cryptology (NETCRYPT), New Delhi, India, pp. 1522-1527 (2025), doi: 10.1109/NETCRYPT65877.2025.11102567.
