Industrial Internet of Things (IIoT) deployments in Industry 5.0 manufacturing settings demand real-time decision making that cloud architectures simply cannot deliver reliably cloud round-trip latencies of 100–300 ms are fundamentally at odds with the sub-10 ms response windows required to prevent equipment damage or injury in safety-critical machinery. This paper presents FogAI, a lightweight, interpretable decision engine that runs directly on the fog computing node, making safety critical decisions in under 5 ms with no cloud dependency whatsoever. The engine combines two complementary mechanisms:a set of five hard-coded safety rules (R1–R5) that guarantee deterministic, instantaneous responses to fault conditions, and a normalised weighted-feature risk scoring model with four adaptive routing rules (SR1–SR4) that intelligently manages subcritical IoT traffic. FogAI is evaluated through a 60-cycle simulation of an industrial CNC machine subjected to seven realistic fault scenarios, ranging from temperature runaway to mechanical vibration shock and full network outages. The results show 75% of all decisions processed locally at a mean fog latency of just 3.1 ms, saving an average of 66.9 ms per cycle compared to routing everything to the cloud, while staying completely operational even during network failures. The entire engine runs in under 0.4 ms of computational overhead with a memory footprint below 8 MB—well within the limits of a Raspberry Pi 4.

Fog Computing; Cloud Computing; DDoSattack; IP Spoofing; OS Fingerprint,Internet of Things (IoT),low latency