Intelligent Cooperative Agents For Ddos Attack Detection and Response In Distributed Clouds

This paper introduces an intelligent, privacy-preserving, and distributed architecture for real-time DDoS detection in distributed cloud environments. The proposed framework integrates a cooperative Multi-Agent System (MAS) with Federated Learning (FL) enabling autonomous monitoring local model training, secure federated aggregation, and real-time inference across geographically distributed nodes. Unlike traditional solutions that rely on centralized processing or static rules, our system supports continuous learning and decentralized anomaly detection with out sharing raw network data. Experiments conducted on physical machines configured demonstrate high robustness achieving 99.8% accuracy, precision, recall, and F1-score, while sustaining a real-time detection rate above 90%. These results highlight the scalability, adaptability, and operational efficiency of the proposed collaborative approach.