Fault tolerance for routers

DDoS attack mitigation using an optimized machine learning approach

Abstract

The increasing complexity of modern networks and the growing frequency of distri-buted denial-of-service (DDoS) attacks demand fault-tolerant mechanisms capable of maintaining service availability. This paper proposes an optimized machine le-arning-based model for real-time DDoS attack detection and mitigation, ensuring continuous operation even under adverse conditions. The system employs the XGBoost algorithm fine-tuned through Particle Swarm Optimization (PSO), which automatically adjusts hyperparameters to achieve higher accuracy and lower infe-rence latency. Experiments conducted in a simulated environment achieved accu-racy above 97% and a significant reduction in response time compared to conven-tional models. Furthermore, the model proved feasible for deployment in embedded devices, sustaining stable performance even under heavy malicious traffic. The ob-tained results validate the effectiveness of the proposed approach and highlight its potential to enhance the resilience and autonomy of critical network infrastructures.