FedXAI for Detecting DDoS on IoT Edge Networks in Federated Learning
Ahmed Asiri, Weiqi Wang, Feng He Wu, Hiep Khac Vo, Shui Yu · 2024
Federated Learning has emerged as an approach to distributed learning that utilizes artificial intelligence (AI) to protect data privacy on edge networks and devices. However, Federated Learning-based Internet of Things (IoT) edge networks can still be vulnerable to distributed denial of service (DDoS) attacks which can negatively influence the operations of Federated Learning models running on these networks. Current methods for detecting DDoS primarily focus on securing devices and data, overlooking model protection. In this paper, we utilize and adapt Federated Explainable AI (FedXAI), a Federated Learning designed with SHapley Additive exPlanations (SHAP) to enhance DDoS detection and interpretation within Federated Learning on IoT networks. FedXAI provides interpretable insights into the models that can be crucial for identifying anomalies indicative of DDoS. Our results show that FedXAI improves DDoS detection, contributing to data and model security with higher accuracy, precision, recall, and F-score than the selected baseline models.