A Network Intrusion Detection Method for Information Systems Using Federated Learning and Improved Transformer

Abstract

A network intrusion detection method for information systems using federated learning and improved transformer is proposed to address the problems of long detection time and low security and accuracy when analyzing massive data in most existing intrusion detection methods. Firstly, a network intrusion detection system is constructed based on a federated learning framework, and the transformer model is used as its universal detection model. Then, the dataset is divided and an improved generative adversarial network is used for data augmentation to generate a new sample set to overcome the influence of minority class samples. At the same time, the new samples are input into the transformer local model for network attack type detection and analysis. Finally, the authors aggregate the detection results of each local model and input them into the Softmax classifier to obtain the final classification prediction results.