Machine Learning Techniques for Cryptographic Attacks: A Systematic Review of Detection Methods

This systematic review investigates the application of machine learning techniques for detecting cryptographic attacks and analyzing system vulnerabilities in cybersecurity environments. A range of models including Naïve Bayes (NB), Decision Trees (DT) C4.5, Random Forest (RF), K-Nearest Neighbors (KNN), Support Vector Machines (SVM), Convolutional Neural Networks (CNN), and hybrid models integrating Particle Swarm Optimization (PSO), were evaluated across 24 peer-reviewed studies. RF and DT models consistently achieved high detection accuracy, with some reporting up to 99.9%, particularly in identifying brute force and Distributed Denial of Service (DDoS) attacks. Hybrid approaches, notably PSO combined with Neural Networks (NNs) or RF, demonstrated enhanced classification precision and recall. The review highlights the importance of model selection, emphasizing trade-offs between detection speed and accuracy. While most evaluations were conducted in simulated environments, the findings offer valuable insights for organizations seeking robust, adaptive, and scalable solutions to combat evolving cryptographic threats. This work serves as a foundational reference for integrating machine learning models into cryptographic systems to improve threat detection and system resilience.