Convolutional Neural Network Approach for Anomaly Pattern Analysis in Information System Log Data

Abstract

The rapid growth of modern information systems has led to the generation of large volumes of log data that record operational activities, system events, and security-related interactions. These logs provide valuable information for monitoring system behaviour and detecting abnormal conditions within complex computing infrastructures. However, manual log inspection and traditional rule-based monitoring approaches are often ineffective in identifying hidden anomaly patterns, particularly in large-scale and dynamic environments. This study proposes a deep learning–based approach using a Convolutional Neural Network (CNN) for anomaly pattern analysis in information system log data. The proposed method focuses on automatically learning hierarchical feature representations from structured log data, enabling the model to capture local patterns and structural relationships among log events without extensive manual feature engineering. Log entries are first processed and transformed into numerical representations suitable for deep learning models. The CNN architecture is then applied to learn discriminative features that distinguish normal system behaviour from anomalous patterns. Experimental evaluation demonstrates that the proposed approach is capable of effectively identifying anomalies within log datasets and provides improved detection capability compared with conventional monitoring techniques. The findings indicate that convolutional neural networks can serve as a reliable method for automated log analysis in large-scale information systems. This research contributes to the development of intelligent monitoring solutions that support early detection of system abnormalities and improve the reliability and security of modern digital infrastructures.