School of Information and Communication Technology, Delta State Polytechnic, Ogwashi-Uku
Academy Journal of ICT
Vol. 1 No. 1 · June 2026
OMOROGIE MICHAEL · NWABUDIKE UJU CYNTHIA
Advanced Persistent Threats (APTs) represent sophisticated, long-term cyber-attacks that evade traditional security measures, posing significant risks to organizations worldwide. This paper explores deep learning-based anomaly detection models as a robust approach to identifying and mitigating APTs. We review key concepts in APTs and anomaly detection, delve into fundamental deep learning architectures such as Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTM) networks, Auto encoders, Generative Adversarial Networks (GANs), and Transformers, and analyse their applications in detecting anomalous behaviours indicative of APTs. Drawing from recent literature, we examine hybrid models that integrate these techniques to enhance accuracy, reduce false positives, and handle complex network data. Case studies from benchmark datasets like UNSW-NB15 and NSL-KDD demonstrate high performance, with accuracies exceeding 98% in some implementations. Challenges such as data scarcity, model interpretability, and computational demands are discussed, alongside future directions including explainable AI and integration with threat intelligence. This work underscores the transformative potential of deep learning in cybersecurity, providing a comprehensive framework for researchers and practitioners to develop resilient detection systems.
OMOROGIE MICHAEL, NWABUDIKE UJU CYNTHIA. (2026). Deep Learning-Based Anomaly Detection Models for Advanced Persistent Threats. Academy Journal of Information and Communication Technology, 1(1).