Machine Learning Techniques for Cyber Threat Detection: A Comparative Study
Esther Chinwe Eze, Fen Danjuma John, Shakirat O. Raji, Grace A. Durotolu · International Journal of Research Publication and Reviews · 2025
The exponential growth of cyber threats in the digital era necessitates advanced detection mechanisms beyond traditional signature-based approaches.This comprehensive study examines and compares various machine learning techniques for cyber threat detection, analyzing their effectiveness, computational efficiency, and practical applicability.We evaluate supervised learning methods, including Support Vector Machines (SVM), Random Forest, and Neural Networks; unsupervised techniques such as clustering and anomaly detection; and emerging deep learning approaches, including Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks.Our comparative analysis encompasses performance metrics across multiple datasets, including network intrusion detection, malware classification, and phishing detection scenarios.Results indicate that ensemble methods achieve the highest accuracy (96.8%) for network intrusion detection, while deep learning approaches demonstrate superior performance in complex pattern recognition tasks with 94.2% accuracy in malware detection.The study also addresses challenges including feature selection, dataset imbalance, adversarial attacks, and real-time processing requirements.Our findings provide practical guidance for cybersecurity professionals in selecting appropriate ML techniques based on specific threat landscapes and operational constraints.