Federated Learning for Internet of Things
The proposed FedDetect learning framework improves the performance by utilizing a local adaptive optimizer and a cross-round learning rate scheduler, and the system efficiency analysis indicates that both end-to-end training time and memory cost are affordable and promising for resource-constrained IoT devices.
Abstract
Federated learning can be a promising solution for enabling IoT cybersecurity\n(i.e., anomaly detection in the IoT environment) while preserving data privacy\nand mitigating the high communication/storage overhead (e.g., high-frequency\ndata from time-series sensors) of centralized over-the-cloud approaches. In\nthis paper, to further push forward this direction with a comprehensive study\nin both algorithm and system design, we build FedIoT platform that contains\nFedDetect algorithm for on-device anomaly data detection and a system design\nfor realistic evaluation of federated learning on IoT devices. Furthermore, the\nproposed FedDetect learning framework improves the performance by utilizing a\nlocal adaptive optimizer (e.g., Adam) and a cross-round learning rate\nscheduler. In a network of realistic IoT devices (Raspberry PI), we evaluate\nFedIoT platform and FedDetect algorithm in both model and system performance.\nOur results demonstrate the efficacy of federated learning in detecting a wider\nrange of attack types occurred at multiple devices. The system efficiency\nanalysis indicates that both end-to-end training time and memory cost are\naffordable and promising for resource-constrained IoT devices. The source code\nis publicly available at https://github.com/FedML-AI/FedIoT.\n