A Transformer and Federated Learning Techniques for Detecting DDoS Attacks in IoT Environments

Distributed Denial-of-Service (DDoS) attacks pose a significant threat to Internet of Things (IoT) networks, emphasizing the importance of deploying effective intrusion detection systems (IDS) to identify and mitigate these attacks. Deep learning (DL) techniques have proven efficient in detecting DDoS attacks in IoT, as reported in numerous studies. Recently, Transformers and Federated Learning (FL) techniques have gained attention in the cybersecurity community for enhancing IDS performance. In this research, we propose a novel, efficient, and lightweight scheme for detecting and mitigating DDoS attacks in IoT networks using transformer-based on FL technology. In our study, we utilize three recently published datasets, namely TON-IoT, CICDDoS2019, and LATAM-DDoS-IoT, to assess the performance of our scheme. These datasets provide invaluable real-world scenarios and provide insightful perspectives on DDoS attacks in IoT environments. By utilizing these contemporary datasets, we aim to enhance the validity and relevance of our research findings, as well as compare the performance of our scheme with existing approaches. Our experimental findings reveal that the proposed scheme achieves a significantly higher accuracy rate compared to state-of-the-art methods. Specifically, our approach achieves an accuracy rate of 99.62% using the TON-IoT dataset, 99.91% using the CICDDoS2019 dataset, and 99.11% using the LATAM-DDoS-IoT dataset.