Trusted Federated Learning Framework for Attack Detection in Edge Industrial Internet of Things

The edge Industrial Internet of Things (IIoT) is highly vulnerable to attacks due to the vast number of connected devices and the lack of security features. Attacks in edge IIoT can lead to significant damage, including data theft, malfunctioning, and privacy breaches. Federated Learning (FL) is a promising approach to detecting attacks by utilizing edge devices' collective intelligence. FL allows devices to collaboratively learn from multiple devices' data without centralized sharing, which preserves data privacy and reduces communication costs. However, FL has vulnerabilities that can compromise model accuracy, privacy, and security. Trusted FL is essential for collaboration among multiple edge IIoT devices while preserving data privacy and security. Trust plays a critical role in the success of FL, as edge IIoT devices must trust that the models are accurately learning and that their data is protected. To address this, we propose an FL framework that uses Federated Averaging (FedAvg) and Convolutional Neural Network (CNN) to detect attacks in edge IIoT. We also propose a mechanism to calculate trust for appropriate edge IIoT device selection by measuring each device's (a.k.a client's) performance during model training. The proposed edge IIoT device selection method, client selection, can fairly select clients for model training and improve trust in the entire system. Although the proposed FL approach does not outperform the centralized ResNet-18 CNN model on experimental analysis, improving its performance can be a promising solution for detecting attacks in edge IIoT.