A Federated Learning Approach to Strengthen IoT-Based Intrusion Detection Against Low-Rate DDoS Attacks
Iftikhar A. Ahmad · 2025
The rising cyber threats have made it necessary to enforce cybersecurity initiatives by organizations. Traditional approaches to intrusion detection systems (IDSs) are mostly centralized and therefore face substantial challenges like data privacy violations, high resource consumption and communication overhead. One promising solution to this issue is federated learning (FL), decentralized machine learning that can enable training models on edge devices to preserve data privacy. Further, FL, a global model is trained with multiple entities without sharing raw data so it can keep their privacy and communication costs are reduced. This is of great value above all in environments as IoT networks where importance of data sensitivity and privacy are crucial. FL achieves this without compromising much on scalability and resource efficiency compared to centralized approaches, which are two of their biggest limitations. Therefore, this paper discusses recent developments in FL for IDS and provides comprehensive analysis along with some initial results of FL in IoT-based intrusion detection. The proposed model achieves 99% accuracy in detecting low-rate DDoS attacks, which outperforms existing approaches.