CAFED-Net: Cross-Adaptive Federated Learning with Dynamic Adversarial Defence for Real-Time Privacy-Preserving and Threat Detection in Distributed IoT Ecosystems

The industrial, urban, and healthcare sectors are now in pressing need of immediate cybersecurity as the distributed network of Internet of Things (IoT) has been rapidly growing in these fields.The centralized system of detecting threats cannot easily fit various IoT settings that span systems that are broad in nature and have high sensitivities on their response rates.New threats are fast emerging against defense systems that are having fixed places and developing not only mobile threats but also spreading to different data areas.Federated Learning (FL) Artificial Intelligence (AI) introduces a training strategy that ensures a solution to data privacy issues, as well as protection against the worst of a data leak that can be disastrous in the event of centralization.An innovative system architecture of FL has been developed to control cross-domain threat detection operations in distributed IoT settings by application of dynamic adaptive adjustment strategies.The system completes local training activities on distributed nodes, merges them with aggregation in FL, and applies an adaptive intelligence-sharing framework.The approach also lays strong models and domain-specific capacities that safeguard data independence.The system proposed will make use of adversarial training, thereby dynamically adapting to dynamically discovered attack vectors as operations progress.Their detecting power and the ability to adapt to the simulation-based assessment, however, prove to be more effective than the baseline models in the circumstances that occur under adversarial drift.In this study, the authors introduce a solution that would allow it to conduct real-time IoT threat detection in a privacyfriendly and scalable manner in response to changing cybersecurity threats.CAFED-Net yielded 87.1 percent accuracy across 12 rounds, 78.6 percent robustness on FGSM, a 15 percent communication cost reduction, and variance less than 2.1 percent among the clients.