Machine Learning-Enabled NTN-Assisted IoT: Mapping the Security Landscape

Non-terrestrial networks (NTNs), encompassing unmanned aerial vehicles (UAVs), low-/high-altitude platforms (LAPs/HAPs), and satellite systems, are increasingly enabling Internet of Things (IoT) applications beyond the limits of terrestrial infrastructure. By combining UAV mobility with satellite and HAP coverage, NTN-assisted IoT supports diverse use cases, including remote sensing, smart cities, intelligent transportation, and emergency response. This paper presents a systematic mapping of machine learning (ML) research in NTN-assisted IoT with a focus on security-related aspects. A keyword co-occurrence analysis of over 2000 publications identifies twelve thematic clusters, including three clusters directly related to security, privacy, and trust. Cluster interconnections are analyzed to reveal dominant research trends and technological dependencies. The first security-focused cluster addresses access control, authentication, privacy preservation, and ML-based intrusion detection in Internet of Drones (IoD) and satellite-enabled systems, while also highlighting feature selection and energy-aware design. The second cluster centers on edge computing-enabled localization and privacy, linking technologies such as GPS, RSSI, LoRaWAN, and differential privacy for smart-city deployments. The third cluster emphasizes blockchain-enabled trust mechanisms, integrating blockchain with aerial image classification, intrusion detection, and secure coordination in IoD environments. Using a connectivity-driven analysis, anchor keywords with strong intra-cluster associations are identified and discussed alongside representative literature. Finally, emerging low-frequency themes are used to outline future directions, including AI-enabled security, trustworthy edge intelligence, autonomous and resilient robotic systems, predictive cyber resilience, and secure cognitive communication for next-generation NTN-assisted IoT.