Privacy-Preserving, Edge-Cloud, and Federated AI for Scalable Decision Support in Critical Applications

Abstract

Scalable decision support in critical applications increasingly requires AI systems that can operate across institutions, devices, cloud services, and privacy-sensitive environments. This structured critical review synthesizes on privacy-preserving, edge-cloud, federated, distributed, and deployment-relevant AI for healthcare, cybersecurity, energy, industrial monitoring, business analytics, agriculture, and human-centered technologies. The review develops an eight-axis taxonomy covering application domain, distributed deployment paradigm, privacy and security function, data modality, architecture family, decision-support function, scalability concern, and evidence role. The corpus indicates that privacy-aware decision support cannot be reduced to a single technique. Instead, scalable AI depends on the alignment of model architecture, data modality, governance needs, communication constraints, and human oversight. Healthcare applications highlight privacy-sensitive diagnosis and screening; cybersecurity and digital-resilience studies emphasize data protection and threat-aware operation; IoT and infrastructure systems foreground latency, sensing, and edge feasibility; and enterprise applications show the importance of auditability and accountable automation. Across domains, federated and edge-cloud systems offer a pathway for distributed intelligence, but they introduce challenges related to non-identical data distributions, model update governance, explanation validity, security exposure, and evidence maturity. Future research should prioritize federated benchmarks, privacy-preserving multimodal learning, communication-efficient architectures, deployment monitoring, and governance-aware reporting standards.