Unsupervised log anomaly detection method based on graph convolutional neural network
CN119066654APending Publication Date: 2024-12-03UNIV OF ELECTRONICS SCI & TECH OF CHINA
0 Cites -1 Cited by
Patent Information
- Application Number
- CN202411166337.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-23
- Publication Date
- 2024-12-03
Smart Images

Figure CN119066654A_ABST
Abstract
The invention discloses an unsupervised log anomaly detection method based on a graph convolutional neural network, and belongs to log anomaly detection. According to the method, detection and identification of potential abnormal behaviors in a system are realized by converting original log data into a causal graph and a feature vector, and the method comprises the following steps: firstly, analyzing original non-structural log data into a structural text template; then coding the structural text template into vector representation by using a pre-training large language model bidirectional encoder representation method; and finally, converting the vector log into a graph structure in which nodes represent log events and edges represent association relationships between the events, then capturing key features of the log on the graph structure based on a graph convolutional neural network, and finally, dividing low-dimensional features of the log into two clusters by using a spectral clustering algorithm, and completing log anomaly detection processing by using the obtained abnormal log cluster. According to the method, by introducing advanced natural language processing and graph convolutional network technologies, the automation degree, accuracy and processing efficiency of log anomaly detection are remarkably improved, the defects in the aspects of rule setting, feature extraction, privacy protection and data demand marking in the prior art are overcome, and efficient and reliable log anomaly detection is achieved.
Need to check novelty before this filing date? Find Prior Art
