Comparative evaluation of Cloud–Edge security architectures for DDoS detection

Abstract

Mobile cloud computing (MCC) improves performance and scalability by offloading tasks from mobile devices to cloud and edge infrastructure. Still, it remains vulnerable to low-speed denial-of-service (LDoS) and distributed denial-of-service (DDoS) attacks. This paper compares centralized, distributed, and hybrid architectures and evaluates their effectiveness using key metrics, including detection accuracy, latency, privacy, scalability, and deployment feasibility. In addition to comparing five representative models based on literature reviews, we conduct experimental evaluations of federated learning-based hybrid models using a recent Internet of Things (IoT) network traffic dataset that reflects modern attack patterns. The results indicate that while centralized models achieve the highest detection accuracy, they suffer from increased latency and reduced privacy; decentralized models improve response speed and privacy but face coordination challenges. Hybrid methods, especially those using federated learning, provide a wellrounded solution by offering strong security, flexibility, and efficient performance, making them ideal for practical use in MCC environments.