Journal of Advances in Mathematics and Computer Science

The increasing virtualisation, programmability, and AI assistance of next-generation network infrastructures simultaneously expand attack surfaces and complicate the reliability of SDN/NFV, cloud/data centre fabrics and 5G/6G. This scoping review maps the use of resilience mechanisms and AI-driven control in adaptive, intelligent, and secure distributed systems. Using a PCC-framed question and a PRISMA-ScR-guided process, we searched the IEEE Xplore database, the Scopus database, and the ACM Digital Library for peer-reviewed studies published between 2015 and 2025. Nineteen empirical studies that met the inclusion criteria were identified, covering SRv6 routing, congestion control and data centre optimisation, WAN traffic engineering, SDN control-plane defence, cloud rollout assurance, and O-RAN/5G RIC/xApp workflows. The evidence was primarily focused on O-RAN/5G implementations and revealed a strong presence of graph-aware learning and reinforcement learning variants, as well as federated learning, supervised learning, and lightweight unsupervised anomaly detection. Reported impacts included measurable operational gains (e.g. reduced inference overhead and reduced flow completion time) and security benefits through mitigation and detection-isolation under adversarial conditions. However, evaluation remains fragmented across metrics and settings. Many studies are limited to emulation or small testbeds, simulations, and there is inconsistent reporting of rollback behaviour, adaptive-attacker robustness and overhead. The evidence suggests that AI can improve reliability and efficiency when automation is limited by telemetry integrity controls, governance measures that can be audited, and safe-fallback mechanisms that stop local failures from spreading system-wide.