Asha Chandi

Аннотация

This study addresses the challenge of detecting Advanced Persistent Threats (APTs) in corporate networks by developing a hybrid multi-modal detection framework. We combine traditional machine learning models, deep learning architectures, and transformer-based models to improve the detection of sophisticated and stealthy cyber threats. A comprehensive dataset, consisting of network traffic and event logs, was processed through rigorous data Volume 10 Issue 02-2025 6 INTERNATIONAL JOURNAL OF COMPUTER SCIENCE & INFORMATION SYSTEM (eISSN: 2536-7919 pISSN: 2536-7900) VOLUME 10 ISSUE 02 Pages: 6-20 preprocessing, feature engineering, and model development. The results show that the hybrid ensemble model, integrating Gradient Boosting and Transformer-based architectures, outperforms all other models, achieving 98.7% accuracy, 98.3% precision, and 97.9% recall, while maintaining a false positive rate below 1%. The model demonstrated exceptional performance in real-world simulations, detecting over 98% of malicious activities. Our findings highlight the importance of combining the strengths of classical and advanced machine learning techniques for effective APT detection and mitigation, providing a reliable, scalable solution for real-time cybersecurity. KEYWORDS Advanced Persistent Threats, APT detection, hybrid models, machine learning, deep learning, transformer-based models, network traffic, cybersecurity, feature engineering, ensemble methods, anomaly detection, real-time deployment.