Research on network security threat detection system based on large language model Optimized by GloVe technology

With the increasing complexity of network attacks, traditional security detection methods have limitations in dealing with new threats and zero-day attacks. In this paper, we propose a large language model network security threat detection system based on BERT. The system takes the network traffic log as the data source, extracts the state features and constructs the feature map vector. At the same time, it combines GloVe word embedding to optimize the feature expression to improve the semantic understanding ability of data. Experimental results show that the proposed method is superior to traditional methods in abnormal traffic detection and unknown attack recognition, and has high accuracy and generalization ability, which provides an effective solution for intelligent network security defense.