Distributed Federated Temporal Behavior Monitoring for Adaptive Cybersecurity Detection in IIoT Environments

Industrial internet of thing (IIoT) systems are intended to have large numbers of connected devices that on an ongoing basis share operational data to enable automation, monitoring, and intelligent control. It is believed that protection of these systems must be reliable, because a cyber-attack against IIoT devices can interrupt production, destroy assets and threaten personnel security for smart manufacturing plants where synchronous intrusion leads that the whole-body conveyor belt stops. This research study presents a deep distributed reinforcement learning framework boosted with metaheuristicguided gated recurrent units for the cybersecurity detection in IIoT networks titled federated deep reinforcement learning with metaheuristic-gated recurrent units. The model represents temporal attack behavior by recurrent gating operations, which are learned by global search-based optimization to achieve the fine-grained sensitivity towards the evolving cyber patterns. More specifically, we are considering that local nodes have an updated policy without receiving raw data, to guarantee privacypreserving operation while adjusting responses to adversary behavior. A federated coordination phase extracts model updates into a unified global security model, providing enhancement in resilience against long-lasting anomalous activities. Experiment results show that the proposed scheme achieves accuracy detection, quick response under dynamic environment, and low communication cost as well more importantly, it is very applicable to different IIoT system infections.