Network Intrusion Detection Based on Hybrid Network Model and Federated Learning

Data is a valuable strategic resource for the development of modern society. However, with the increasingly complex network environment, privacy leaks and malicious attacks emerge in endlessly. For example, blockchain has also begun to become a new outlet for network black production, which poses a huge security threat to cryptocurrency. In this paper, we propose a hybrid network model (Cb Net), which uses Convolutional Neural Networks (CNN) and Bidirectional recurrent neural networks (BiGRU) to fully extract the space-time characteristics of network data traffic. Then, we propose an intrusion detection method (FLD), which introduces federated learning to collect traffic data from different network institutions, analyze network traffic and identify network attacks. We have fully evaluated the performance of the proposed model and method on the public dataset NSL-KDD. Experiments show that the proposed hybrid network model can achieve high detection accuracy, and the FLD method can effectively identify network attacks on the premise of ensuring the privacy of the local data of the users involved, and its performance is better than other methods.