An efficient infiltration and denial of service detection using dynamic weighted aggregation federated learning
R.S. Ramya, K.L. Sanjana, K.M. Sharan, G.V.S. Dinesh Reddy, N. Vinod, K. R. Venugopal · 2024
The continuous evolution of intrusion threats in cybersecurity calls for the development of new and robust detection techniques. This paper proposes a novel method of Network intrusion detection system using Federated learning in which the proposed model is capable of detecting attacks such as DoS, DDoS, Infiltration and other attacks. CNN networks are used to create the suggested model, the proposed system presents a novel idea called Dynamic Weighted Aggregation Federated Learning (DWAFL). Furthermore, in contrast to a traditional intrusion detection system based on federated learning, the proposed method has dynamically integrated weighting and filtering algorithms for local models. DWAFL can identify network intrusions more effectively and with less communication overhead. We provide the comprehensive design for DWAFL, and our test findings show that DWAFL can maintain data privacy while achieving good detection performance with a low network communication overhead.