%0 Journal Article
%T 基于GRU-TextCNN的日志序列异常检测方法<br>GRU-TextCNN-Based Anomaly Detection Method for Log Sequences
%A 胡标
%A 徐克
%A 简文
%A 姜咏绮
%A 刘军
%J Computer Science and Application
%P 1006-1018
%@ 2161-881X
%D 2023
%I Hans Publishing
%R 10.12677/CSA.2023.135098
%X <span style=\"font-size:14px;\">系统详细记录着系统的运行情况和事件，因此系统维护人员常常基于日志对系统状态进行分析，判断系统有无出现异常，以此更好地维护系统。由于现代系统日志数据大规模增加，传统的日志异常检测方法已经不适用现代系统日志。本文基于深度学习技术，提出了一种基于GRU-TextCNN的日志异常检测方法。该方法首先通过预处理将日志处理成日志语句，然后利用SBERT模型将日志语句转换成相应的句向量，随后利用滑动窗口提取日志序列，最后利用本文提出的基于GRU-TextCNN的日志序列异常检测模型检测日志序列。通过在两个数据集上的实验结果表明，该方法能够有效检测出日志序列异常。</span><br />
<span style=\"font-size:14px;\">The system meticulously records its operations and events, allowing system maintenance personnel to frequently analyze its status based on log data. This analysis is crucial for determining any abnormalities and ensuring optimal system maintenance. However, with the immense growth in modern system log data, traditional log anomaly detection methods have become inadequate for contemporary systems. In this paper, we introduce a deep learning-based log anomaly detection method utilizing GRU-TextCNN. This method begins by preprocessing logs into log statements, followed by converting these statements into corresponding sentence vectors using the SBERT model. Next, log sequences are extracted through sliding windows, and the log sequence anomaly detection model, based on GRU-TextCNN, is applied. Experimental results from two datasets demonstrate the effectiveness of this method in detecting log sequence anomalies.</span>
%K 日志异常检测，深度学习，GRU，TextCNN<br>Log Anomaly Detection
%K Deep Learning
%K GRU
%K TextCNN
%U http://www.hanspub.org/journal/PaperInformation.aspx?PaperID=65923