Federated edge AI framework for resilient DDoS defense in edge computing

Edge computing enables low-latency and bandwidth-efficient processing near end devices and serves as a critical component of modern digital infrastructure. However, its distributed and resource-constrained architecture makes edge networks highly susceptible to distributed denial-of-service (DDoS) attacks, which can severely compromise network performance and cyber resilience. This paper employs the NS-3 simulator to model DDoS attacks in an edge computing environment and evaluate their impact using FlowMonitor and NetAnim. Key metrics, including packet delivery ratio, latency, throughput, and packet loss, are analyzed to assess network performance under attack conditions. Experimental results reveal significant degradation in service quality, with packet delivery ratios dropping below 60% and latency increasing substantially. To enhance network resilience, a defence-in-depth framework integrating rate limiting, IP-based access control, and threshold-based anomaly detection is implemented at the gateway. The proposed mitigation mechanism restores packet delivery ratios to approximately 90% while considerably reducing latency and packet loss. Furthermore, the study investigates the role of privacy-preserving federated learning and edge artificial intelligence in accurate attack detection using quantized LightGBM and shallow autoencoders. The proposed lightweight Edge AI model achieves an accuracy of 98.6%, exhibits a minimum detection latency of 41 ms, and utilizes only 38% of the available processing capacity. Additionally, the study examines the implications of emerging quantum computing capabilities on conventional DDoS defence mechanisms and highlights the need for quantum-resilient cybersecurity strategies.