A Novel FedLLM Intrusion Detection Frameworkfor Privacy-Preserving Security in IoT EnabledSmart City Network

The proliferation of Internet of Things (IoT) devices in Smart City infrastructureshas introduced unprecedented opportunities for automation, healthcare,and transportation, while simultaneously exposing the ecosystem to complex andmulti-dimensional cyberattacks. Conventional intrusion detection systems (IDS)either lack semantic reasoning, ignore topological dependencies, or fail to preservedata privacy across heterogeneous networks. To address these challenges, we proposea novel FedLLM-based hybrid IDS framework that integrates large languagemodel (LLM)-driven semantic encoding, feature-based traffic representation, andgraph neural network (GNN) embeddings to capture diverse attack patterns. Theproposed architecture leverages a lightweight Transformer-based fusion detectorfor resource-efficient anomaly detection, while employing federated learningwith differential privacy to enable secure collaborative training without rawdata exchange. Experimental evaluation on benchmark IoT datasets, includingCIC-IDS2017, demonstrates that the framework achieves superior accuracy(98.5%), macro F1-score (97.0%), and AUROC (99.9%) compared to state-ofthe-art baselines. Furthermore, the model exhibits lightweight capacity with only0.61 MB size and 158k parameters, enabling deployment on edge devices withminimal latency. By providing human-interpretable alerts through LLM explanations,the framework ensures both operational reliability and transparency inSmart City environments. The proposed framework establishes state-of-the-art1 performance, combining privacy preservation, high accuracy, and scalability forintrusion detection in IoT-based Smart City environments.