CRNN-Based Automated Log Anomaly Detection for Large-Scale Cloud Environments

Abstract

With the development of cloud computing in recent years, the quantity and types of log data have rapidly increased, making issues related to timely anomaly detection and operational security increasingly severe. The purpose of this paper is to construct a large-scale solution for automatic anomaly detection in cloud-generated log streams with full climate protection. Combining Convolutional Recurrent Neural Networks (CRNN) for integrating advanced feature engineering and deep learning to extract local spatial features and long-term temporal dependencies from heterogeneous log data. Rigorous experiments were conducted on a real-world dataset of over 80 million log entries from multiple cloud sources. The proposed model achieved an F1 score of 0.88 and an AUC of 0.974, surpassing previous baseline performances such as PCA, traditional machine learning, and independent deep learning methods. According to comprehensive experiments, CRNN can efficiently handle large amounts of data, is less sensitive to noise and changes in log formats, and performs well in cross-domain situations. This model meets the current requirements of cloud environments and can provide near-real-time detection and adaptation in distributed systems. The CRNN-based framework can automatically and reliably address log anomaly detection issues, providing strong support for future research and development in cloud security monitoring and intelligent event response.