Intrusion detection method and device based on local and global attention mechanisms and medium

Abstract

The application relates to the field of log detection, in particular to an intrusion detection method and device based on local and global attention mechanisms and a medium, the method comprising the following steps: acquiring system logs to be detected; extracting features of the system logs to be detected; extracting local attention features of the features by using a local feature attention mechanism; extracting global attention features of the features by using a global feature attention mechanism; and performing sequence prediction according to the local attention features and the global attention features to obtain an intrusion detection result of the system logs to be detected. By implementing the application, the local features and the global features are extracted, on one hand, the attention mechanism based on the close-range local features focuses on constructing the system log character-level local features, on the other hand, the attention mechanism based on the long-distance global features focuses on capturing the long-distance context features, the hidden feature information in the logs is learned more comprehensively, the accuracy of the intrusion detection is further improved, and the intrusion detection method is more effective.

Images