Advanced Detection of DDoS Attacks Using Hybrid Deep Learning Models and Federated Learning

Distributed Denial of Service (DDoS) attacks are an increasingly serious threat in the realm of cybersecurity, requiring sophisticated methods for detection and response. This paper presents an innovative machine learning framework that combines hybrid deep learning models, graph neural networks, and federated learning to enable real-time detection and analysis of these attacks. By utilizing advanced algorithms and refined feature engineering on large-scale datasets such as NSL-KDD and CICIDS2017, this research focuses on efficiently monitoring and analyzing network traffic. What sets this framework apart is its use of graph-based methods to improve anomaly detection in network traffic, alongside the application of federated learning to facilitate decentralized model training while safeguarding data privacy. The proposed system achieves notable advancements in detection accuracy, with results showing 98.7% accuracy on the NSL-KDD dataset and 97.5% on the CICIDS2017 dataset. Furthermore, the framework is designed to be both adaptable and responsive, providing faster detection times than traditional approaches, thereby equipping network administrators with a powerful tool to proactively manage DDoS threats.