A Federated Flower Learning-Empowered Intrusion Detection System for Vehicle 6G Fog Networks

The evolution of intelligent transportation systems in the era of 6G has introduced new challenges in cybersecurity, particularly for vehicular networks interconnected through fog and edge nodes. This research proposes a Federated Flower Learning-Empowered Intrusion Detection System (IDS) for Ve-hicle 6G Fog Networks using the TON_loT dataset, aiming to achieve high-accuracy, low-latency attack detection while preserving data privacy. The TON_loT dataset, collected from the Cyber Range and loT Labs at UNSW Canberra, offers a diverse range of telemetry and system data from Internet of Things (loT) sensors, fog operating systems, and network traffic. The dataset encompasses both benign and malicious behaviors, including attacks such as DoS, DDoS, and ransomware, within a realistic Industry 4.0 testbed comprising virtualized fog and loT environments. The proposed IDS leverages a Convolutional Neural Network (CNN) trained on this rich dataset to extract spatial features and detect intrusion patterns. To maintain data locality and privacy across distributed vehicular sensor nodes, a federated learning architecture using the Flower framework is implemented. In this architecture, each vehicle acts as a federated client, locally training the IDS model on its sensor data and sharing only model updates with a central aggregator. This eliminates the need to transfer raw sensitive data while ensuring collaborative model refinement. The system is designed to monitor in-vehicle sensor data in real-time and raise alerts in the presence of anomalies or malicious activity. The federated CNN - based IDS achieves high detection accuracy and minimises end-to-end latency, ensuring timely responses in critical vehicle control scenarios. Simulation results show that the proposed system reduces bandwidth consumption and response delays compared to traditional cloud-based IDS solutions and identifies the highest security risk from both known and unknown malicious threats in distributed vehicle 6G networks.