Abstract

Next-generation cellular networks including 5G and emerging 6G infrastructures are transforming global connectivity with ultra-low latency communication, massive device scaling, and intelligent edge computing. However, these advancements significantly broaden the cyberattack surface, creating critical vulnerabilities in network slicing, software-defined architectures, multi-access edge computing (MEC), and heterogeneous Internet of Things (IoT) deployments. To address these challenges, this manuscript proposes MobileShield AI, a novel autonomous security framework that integrates deep learning-driven threat analytics, zero-trust access control, and behavioral anomaly detection across cellular network layers. MobileShield AI leverages federated learning to ensure privacy-preserving intelligence sharing among distributed network entities while enabling real-time threat hunting at the edge. The system architecture incorporates dynamic risk scoring, adaptive policy reinforcement, and blockchain-aided trust verification providing strong resilience against evolving attacks such as DDoS, signaling storms, identity spoofing, and rogue base stations. Experimental validation through simulations demonstrates that the framework improves intrusion detection accuracy, reduces latency in security response, and enhances overall network integrity compared to conventional models. MobileShield AI establishes a scalable and proactive security paradigm suitable for future hyper-connected environments, contributing to secure operational continuity across telecom ecosystems. The proposed solution sets foundational directions for integrating AI-driven defense into standards and deployments of upcoming cellular infrastructures.