Author(s)

Shaopei Su, Yan Zhang

Corresponding Author

Shaopei Su

Abstract

This article focuses on anomaly detection of network logs, aiming at building an efficient anomaly detection model and improving its performance under the background of increasingly complex network systems and limited traditional detection methods. Through in-depth analysis of related concepts of network logs, artificial intelligence (AI) technology and anomaly detection principle, a hierarchical architecture design model is adopted, and the long-term and short-term memory network (LSTM) algorithm based on deep learning (DL) is applied, combined with data cleaning, over-sampling and parameter optimization strategies. The experimental results show that the model with improved performance is significantly superior to the unoptimized model in terms of accuracy, precision, recall and F1 value, and the detection accuracy of different types of abnormal logs is also greatly improved. The research shows that the model and performance improvement method are effective and provide a reliable scheme for anomaly detection of network logs, but it still needs to be continuously optimized to adapt to the dynamic network environment.

Keywords

Artificial Intelligence; Network Log; Anomaly Detection Model; Performance Improvement; Network Security