A Lightweight Federated Transfer Learning Framework for DDoS Attacks Detection in the Internet of Medical Things (IoMT)
Year: 2026
Type: conference-paper
Abstract: The rapid incorporation of the Internet of Medical Things (IoMT) has improved healthcare services by providing real-time monitoring of patients and advanced medical applications. The growing interconnectivity of IoMT devices has made healthcare networks vulnerable to Distributed Denial-of-Service (DDoS) attacks. These attacks have the potential to disruptessential medical services and jeopardize patient safety. Traditional centralized intrusion detection techniques suffer from privacy concerns, limited scalability, and high communication overhead. To overcome these challenges, we proposes FedTL-DDoS, a lightweight Federated Transfer Learning framework for the detecting DDoS attacks in IoMT environments. The proposed framework incorporates federated learning for privacy-preserving training of distributed models, transfer learning to improve detection performance with limited labelled data, and a lightweight 1D-CNN model optimized through pruning, quantization, and gradient compression for resource-constrained IoMT devices. A novel adaptive federated aggregation strategy significantly enhances the robustness to non-IID data distributions and malicious updates. The Experimental findings on benchmark IoMT datasets show that the suggested framework attains high detection accuracy with reduced model size, lower communication overhead, and faster inference compared with existing centralized and federated techniques. The results demonstrate that FedTL-DDoS is an effective, scalable, and privacy-preserving approach for securing next-generation IoMT networks against DDoS attacks.
Authors: Tejinder Sharma, Bharti Sharma
Institution: Maharishi Markandeshwar University, Mullana
Language: English
Subfield: Computer Networks and Communications
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7225757749