Federated Zero Trust Multi-Agent System for DDoS Defense with Blockchain

Abstract

The spread of distributed network environments has led to a rise in the risk of Distributed Denial-of-Service (DDoS) attacks, which interrupt services by flood-ing network resources. Centralized traditional intrusion detection systems have various shortcomings such as issues with scalability, communication overhead, and privacy concerns of data. In order to tackle these problems, the present work puts forward a Federated Learning-based DDoS detection method that ensures privacy and security. The proposed system uses a combined CNN-BiLSTM model to accurately identify spatial and temporal features in network traffic. An RL-DQN, which stands for reinforce men learning-based feature selection, method is implemented to select the most relevant features there by reducing the time complexity and at the same time ensuring high detection accuracy. Besides, security has been improved by the incorporation of Zero Trust Architecture for continuous node verification, Byzantine-resilient aggregation methods to prevent adversarial updates, and homomorphic encryption for secure communication of model parameters. Furthermore, a Multi-Agent System permits adaptive decision- making, while blockchain technology provides transparency and the immutability of system logs. Based on the CIC-DDoS2019 dataset, the results show that the proposed framework can attain high precision, be scalable, and resilient in ever- changing and hostile environments which makes it a viable solution for the next generation of distributed cyber security systems.