Federated Few-Shot Learning for Robust and Privacy-Driven Network Intrusion Detection in IoT

With the rapid growth of IoT networks, traditional centralized Network Intrusion Detection Systems (NIDS) face challenges in maintaining data privacy while processing large volumes of data. Federated Learning (FL) enhances privacy by enabling distributed learning without sharing raw data. However, its effectiveness is limited by the scarcity of labeled attack data. To address this, we propose a Federated Few-Shot Learning (FFSL) framework for IoT network intrusion detection. The framework integrates Few-Shot Learning (FSL) to improve NIDS generalization from a few labeled examples. Meanwhile, FL facilitates collaborative training across distributed IoT devices while preserving privacy. The proposed approach employs a metric-based method combined with Long Short-Term Memory (LSTM) networks to capture temporal dependencies in network traffic, enhancing detection accuracy. The proposed technique is evaluated on the Telemetry of services, Operating systems, and Network traffic dataset for the IoT (ToN_IoT) dataset, the method achieves higher precision, recall, and F1-score compared to state-of-the-art techniques while ensuring computational efficiency and data privacy.