AI-Driven Security Frameworks for IoT Devices in 5G Edge Environments: A Survey

Abstract

The convergence of the Internet of Things (IoT), fifth-generation (5G) networks, and edge computing is revolutionizing connectivity, enabling massive device integration, ultra-low latency, and real-time applications. However, this paradigm shift introduces an expanded attack surface, exposing IoT ecosystems to diverse and sophisticated cyber threats across device, network, and edge levels. Traditional security mechanisms are often inadequate due to scalability constraints, resource limitations, and the need for adaptive responses in dynamic environments. Artificial Intelligence (AI), encompassing machine learning (ML), deep learning (DL), reinforcement learning, and federated learning, offers promising solutions by enabling intelligent, adaptive, and decentralized defense mechanisms. This survey provides a comprehensive analysis of AI-driven security frameworks for IoT in 5G edge environments. We develop a taxonomy of AI-enabled approaches, including intrusion detection and prevention, malware and botnet detection, privacy-preserving learning, intelligent access control, and blockchain-AI hybrid models. A comparative analysis of recent frameworks (2021–2025) highlights their AI techniques, threat target, datasets, latency suitability, key metrics reported and scalability notes, while underscoring the gaps in scalability, real-time detection, and privacy assurance. Furthermore, we discuss benchmark datasets, emerging evaluation metrics, and identify pressing research challenges such as lightweight AI models, explainable AI, and quantum-era security considerations. Finally, the paper envisions the future of adaptive, self-healing IoT security in 6G and beyond, emphasizing the integration of blockchain, neuromorphic computing, and edge AI.