FedDDoS: An Efficient Federated Learning-based DDoS Attacks Classification in SDN-Enabled IIoT Networks
Ahmad Zainudin, Rubina Akter, Dong-Seong Kim, Jae‐Min Lee
Abstract
Independent distribution systems are made possible by Industry 4.0, and these systems produce heterogeneous data that is vulnerable to cyberattacks. The Distributed Denial of Service (DDoS) attack is a typical contemporary cyber threat that disables a target server by flooding it with malicious traffic. In this research, a deep-federated learning-based decentralized DDoS classification method enables independent clients to train local data while maintaining each industrial agent's data privacy. This framework applies a filter-based Pearson correlation coefficient (PCC) feature selection technique for selecting potential features to reduce complexity and improve the model performance. The proposed model has been evaluated with the recent DDoS attacks dataset, CICDDoS2019, and achieves great accuracy of 98.37% with a computational time of 3.917 ms.