Journal Article

AI-Enhanced Intrusion Detection for UAV Systems: A Taxonomy and Comparative Review

Mahmoud, Ashraf Sharif, Sheltami, Tarek Rahil, Islam, MD Sakibul

2025

Overview

The diverse usage of Unmanned Aerial Vehicles (UAVs) across commercial, military, and civil domains has significantly heightened the need for robust cybersecurity mechanisms. Given their reliance on wireless communications, real-time control systems, and sensor integration, UAVs are highly susceptible to cyber intrusions that can disrupt missions, compromise data integrity, or cause physical harm. This paper presents a comprehensive literature review of Intrusion Detection Systems (IDSs) that leverage artificial intelligence (AI) to enhance the security of UAV and UAV swarm environments. Through rigorous analysis of recent peer-reviewed publications, we have examined the studies in terms of AI model algorithm, dataset origin, deployment mode: centralized, distributed or federated. The classification also includes the detection strategy: online versus offline. Results show a dominant preference for centralized, supervised learning using standard datasets such as CICIDS2017, NSL-KDD, and KDDCup99, limiting applicability to real UAV operations. Deep learning (DL) methods, particularly Convolutional Neural Networks (CNNs), Long Short-term Memory (LSTM), and Autoencoders (AEs), demonstrate strong detection accuracy, but often under ideal conditions, lacking resilience to zero-day attacks and real-time constraints. Notably, emerging trends point to lightweight IDS models and federated learning frameworks for scalable, privacy-preserving solutions in UAV swarms. This review underscores key research gaps, including the scarcity of real UAV datasets, the absence of standardized benchmarks, and minimal exploration of lightweight detection schemes, offering a foundation for advancing secure UAV systems.

• Publisher: MDPI AG
• Subject: Artificial intelligence / Artificial neural networks / Control systems / Cybersecurity / Cyberterrorism / Data security / Datasets / Deep learning / deep learning IDS / Digital libraries / Drone aircraft / Federated learning / Intrusion detection systems / Literature reviews / Machine learning / machine learning IDS / Methods / Neural networks / Prevention / Real time / reinforcement learning IDS / Supervised learning / Taxonomy / UAV dataset / UAV intrusion detection / UAV network attacks / Unmanned aerial vehicles / Wireless communications